IGC Evidence Paper Cities

Gharad Bryan Edward Glaeser Nick Tsivanidis

IGC evidence paper Cities

DIRECTED BY FUNDED BY

EVIDENCE PAPER CITIES / MAY 2021 Table of Contents

1 Introduction 3

2 Agglomeration economies 6 a. Is there economic opportunity in developing-country cities? 6

b. Empirical estimates of agglomeration economies 8

c. Can developing-country cities become more productive? 13

d. How can the economic benefits of cities be more widely shared? 15

3 Infrastructure, institutions, incentives, and density’s downsides 17 a. What are the social costs of urban contagion, congestion, and crime? 18

b. Incentives and behavioural change 20

c. Estimating the social benefits of infrastructure 21

d. Institutional reform and public capacity 23

e. Housing, planning, public finance, and water and waste management 24

f. Cities and climate change 27

4 The structural approach to transportation and land use 30 a. The basic form of structural urban models 30

b. What’s different in developing countries? 32

c. Better parameter estimates to make structural models more useful 35

5 Conclusion 39 6 References 40

Cover image: Babatunde Olajide

EVIDENCE PAPER / CITIES 2 1 Introduction

The United Nations forecasts that “Africa’s urban population is likely to nearly triple between 2018 and 2050”. Together, Africa and India account for almost two thirds of the projected growth in the world’s urban population from 4.2 billion in 2018 to 6.7 billion in 2050. The urbanisation of our planet’s poorer countries is one of the most important phenomena of the twenty-first century and a critical component of structural change. Yet, our intellectual tools for dealing with the great challenges of developing-country cities remain under- developed. In this paper, we survey the economics of developing-country cities and try to make the case that development economists should spend more of their time thinking about and working in cities and urban economists should spend more of their time thinking about and working in developing countries. Throughout most of history, agricultural prosperity and transportation infrastructure generally preceded mass urbanisation, because agricultural surpluses were needed to feed city dwellers. Today, much of the developing world is urbanising at lower levels of income and with less developed governance than America enjoyed when it became a majority-urban nation in 1920. Consequently, developing-country cities must face the ubiquitous challenges of urban life with far fewer resources. The difficulties of developing-country urbanisation only increase when national leaders seek to stymie urbanisation for some reason—perhaps out of a fear that cities will enable the mobilisation of political opposition. The study of developing-country cities provides a window into topics at the heart of economics. Cities are home to externalities, both positive and negative; they can represent the highest and lowest points of human behaviour. The knowledge-based growth described by Paul Romer and Robert Lucas takes tangible form in urban areas. Pigouvian problems, such as traffic congestion and contagious disease, become hyper-charged in the extreme densities of poorer cities. In this paper, we divide the field of urban development economics into three broad categories: agglomeration economies, density’s downsides, and spatial models of transportation and housing. The central question of agglomeration economics is if cities themselves actually increase productivity or if the observed relationship between density and earnings represents the selection of more skilled people into cities, or if there are some omitted variables that both attract people and make them wealthier. The growing literature on urban development appears to confirm the positive effects of urbanisation on earnings that have been found in the wealthy world (Chauvin et al. 2017). Randomised control trials that induce migrants to come to cities have provided some of the most compelling evidence supporting the hypothesis that density increases earnings (Bryan, Chowdhury, and Mobarak 2014).

EVIDENCE PAPER / CITIES 3 Yet, there is also evidence suggesting that city slums are home to millions of people who have lived in urban areas for decades and remain poor (Marx, Stoker, and Suri 2013). Resolving the question of whether developing-country slums are dead ends or offer pathways to prosperity remains central to research on developing world agglomeration. There is also a need for research that uncovers the means of improving developing-country cities’ productivity, as well as how to spread the benefits of urban productivity more widely. Urban proximity enables poorer workers to connect with employers, but it also enables the spread of disease and the perpetration of crimes. Western cities were known for epidemics—cholera, typhoid, influenza—until the early 20th century, and water-borne illnesses remain a serious challenge in the world’s poorest cities. Even relatively rich New York had high murder rates through the early 1990s, and homicide continues to bedevil many Latin American and African cities today. Demand for urban density can also collide with a limited housing supply and make living space unaffordable. Density can also slow actual mobility; as the automobile became more widely used throughout the twentieth century, spreading even to cities with limited public capacity, traffic congestion became severe in many urban areas. Today, workers in highly congested cities, such as Jakarta and Sao Paulo, can often face commutes that exceed one hour. These types of challenges posed by density create scope for research and public policy action that can potentially make developing-country cities more liveable for all their inhabitants. Economists are increasingly analysing the roles that incentives, infrastructure, and institutions can play in moderating urban crime, traffic congestion, and disease in developing-country cities. High levels of homicide in many poor world cities has been linked to extremely low probabilities of arrest and punishment. A large and growing literature is examining how certain institutions, such as public-private partnerships, impact road maintenance and demand management. One major finding of this literature is that weak public institutions do not imply better performance by private institutions; such private providers of public services often have incentives to subvert the government that is allegedly overseeing them (Engel, Fischer, and Galetovic 2014). In many cases in the developing world, where governments are ineffective or mired in complex bureaucratic systems, the road to a solution can be less clear than it initially appeared. These complex environments must be considered when proposing any public policy changes. For example, analysing the impact of land use in a city requires fully fledged spatial models that can assess the full equilibrium implications of building up one area of the city. Similarly, large-scale changes in transportation infrastructure may have impacts that ripple throughout a metropolitan area. Section IV of this paper focuses on the growing subfield of developing structural spatial models that can use empirically estimated parameters and forecast the city-wide impact of policy changes.

EVIDENCE PAPER / CITIES 4 While many development economists have been rightfully enthusias- tic about the scientific precision generated by randomised control trials, cities are complex systems, and many urban problems cannot be addressed solely through performing research interventions that can be randomised at the individual. The structural approach to urban economics typically embeds a series of optimisation problems, including the locations and employment decisions of people and firms, and developers’ decisions about construction. These models’ parameters are then estimated directly from the data or by using other sources of information, including randomised control trials. Different policy choices can then be simulated using these parameter estimates. These models are just starting to be applied to contemporary policy challenges, but structural spatial models seem well-suited for land use and transportation decisions in developing-country cities. The future of the developing countries is urban, and this fact generates both challenges and opportunities. The research that we now discuss represents the beginnings of a robust literature on developing-country cities. There is every reason to believe that this literature will continue to grow and that it will provide fascinating policy-relevant results.

EVIDENCE PAPER / CITIES 5 2 Agglomeration economies

Should national governments work to promote or restrain the process of urbanisation? The case for active regulation of spatial policy depends on many factors, particularly whether cities actually enhance productivity or are simply correlated with it. Empirical estimates of the true causal magnitude of agglomeration economies are therefore crucial elements in this most basic urban policy question. Figure 1 documents two remarkable facts. The first panel plots, at the country level, the correlation between non-agricultural labour share and the log of output per worker in both agricultural and non-agricultural fields. The poorest countries in the world are predominantly rural and agricultural; these can be found on the left side of the figure. Not only are developing countries relatively worse at agriculture, but most of their workers labour on farms (Vollrath 2014), implying a productivity disconnect. The second panel shows the correlation between urbanisation in 1960 and growth between 1960 and 2010 among a sample of poor countries in 1960. To paraphrase Lucas (1988), these figures suggest enormous possibilities. Is there something that Malawi could do, some action that its government could take, that would allow the 75% of its workers that work in rural areas in agriculture to access the productivity levels of its non-agricultural, more urban workers, increasing their productivity above that of agricultural workers in, for example, Great Britain? In this section, we will try to understand whether these possibilities are real or whether higher urban productivity might just reflect the selection of more skilled people or better firms into cities.1

A Is there economic opportunity in developing-country cities?

Figure 1a suggests that if more people lived in cities in the developing world, productivity and wages could be higher. In this section, we review three classes of theories that might explain this phenomenon—each consistent with the facts, but having different implications for whether these opportunities are real or merely illusory. The first model is that more able people choose to live in cities, which would occur if people who have an absolute ability advantage also enjoy a

1 Gollin, Lagakos, and Waugh (2014) investigate and reject the hypothesis that the urban productivity advantages suggested by Figure 1a are purely measurement error.

EVIDENCE PAPER / CITIES 6 FIGURE 1A Cross Country Productivity Gaps

FIGURE 1B Urbanisation and growth

EVIDENCE PAPER / CITIES 7 comparative advantage of producing in cities.2 A second possibility is that the urban wage premium is real, but that amenity losses or high housing prices ensure there is no overall welfare benefit from increasing urbanisation.3 That model still suggests an urban productivity premium—after all, why else would private sector employers be willing to pay higher wages in cities? However, it does not indicate a welfare premium for rural-to-urban migrants. A third model is that city size generates positive externalities that might be static or dynamic in nature. Static externalities might occur because a larger market size encourages the entry of new product varieties, as in Krugman (1991). Dynamic externalities might occur if cities spread ideas and speed up the right kind of technological progress, as suggested by Lucas (1988). If the urban wage premium just represents omitted individual characteristics, then there is little reason to think that moving to cities will make people or the country as a whole more productive. If it represents place-specific assets, then moving to that area will make people more productive but will not have any positive effect on overall regional welfare. If the urban wage premium represents local externalities, then workers’ relocation to the area may generate benefits for existing residents or the country as a whole. These externalities typically represent market failures.

B Empirical estimates of agglomeration economies

Urban workers do typically earn more, but does this represent a true effect of place or merely the selection of more able people into cities? The simplest and most common approach to measuring the economic benefits of agglomeration is to run an individual-level regression where earnings are regressed on individual characteristics, such as age and education, and local characteristics, such as area density or total population agglomeration. Within the US, such

2 Lagakos and Waugh (2013) note that if absolute and comparative advantage are independent, then a small wedge or friction can lead to large differences in productivity between rural and urban dwellers. Bryan and Morten (2019) use a structural model that assumes that absolute and comparative advantage are uncorrelated and Indonesian data to estimate the speed with which average wages drop with movement across space. They find that the elasticity of average wages with respect to the proportion of an original population moving is around -0.039. In their setting, this implies that despite large spatial wage differences, moving people across space offers only moderate gains. 3 In the classic Harris and Todaro model, urban unemployment is higher than rural unemployment, but urban workers earn higher wages if they are lucky enough to be employed.

EVIDENCE PAPER / CITIES 8 estimates typically yield a coefficient of .05 when the logarithm of wages are regressed on the logarithm of population (Ciccone and Hall 1996, Glaeser and Gottlieb 2008, Ahlfeldt and Pietrostefani 2019), meaning that wages increase by about 5% when total population or density doubles.4 Chauvin et al. (2017) perform three comparable exercises for Brazil, China, and India, finding large effects of density on earnings, especially in India and China. Young (2013) uses Demographic and Health Survey data to construct consumption equivalents of education, documenting large differences in consumption levels between rural and urban areas in a sample of 65 countries that includes many of the poorest in the world. His results show that the urban/rural wage gap accounts for about 40% of within-country inequality in his sample, but he also notes strong sorting on observable characteristics, which suggests that sorting on unobservables may also be important. Gollin, Kirchberger, and Lagakos (2017) document large consumption differences across density levels in twenty African countries. To address the problem of selection on unob- servable attributes, researchers have increasingly re- lied on migration, natural experiments, and even ran- Workers who move domised control trials to estimate the treatment effect to large urban areas of place on earnings. Glaeser and Mare (2001) found- experience faster ed the literature that estimates the urban wage premium by looking at the wage gains experienced by ur- wage growth in the ban-rural migrants. The key identifying assumption is years following. that unmeasured worker ability doesn’t change over time, or at least that changes in unobserved worker ability are not correlated with moving across space. Glaeser and Mare (2001) find that workers who move to large urban areas experience faster wage growth in the years following, which is compatible with the view that cities enable human capital accumulation. De la Roca and Puga (2017) use administrative data that enables them to follow the wage patterns of almost all workers in Spain as they move across geographies. They also find that workers who come to large cities, like Barcelona and Madrid, experience wage gains over time.5 In developing countries, Hicks et al. (2017) use panel data from Kenya and Indonesia to present fixed effect estimates of the urban rural wage gap. Their fixed effect estimates show that urban workers in Indonesia earn 2.8% more per hour while urban Kenyans earn 26% more.

4 Combes et al. (2010) is a particularly effective paper estimating agglomeration effects in France. The authors control for firm characteristics and use soil characteristics as an exogenous source of variation for density. 5 Chetty and Hendren (2018) use income tax data in the U.S. to look at families who move across areas and establish the impact of place on economic opportunity.

EVIDENCE PAPER / CITIES 9 Though limited by the lack of panel data sets, other researchers have produced work surveying rural-urban migrants who moved first to impover- ished neighbourhoods. Perlman (2010) starts with an initial sample of favela-dwellers in Brazil in 1969 and looks at the outcomes for their children and grandchildren. She finds that while 72% of the first generation were illiterate and 94% worked in manual jobs, by 2001, only 6% of their children were illiterate, and 63% held manual jobs. By the time their grandchildren come of age to work, only 39% hold manual jobs. Alesina et al. (2019) similarly find that intergenerational upward mobility is related to urbanisation in Africa. By contrast, Marx, Stoker, and Suri (2013) examine a cross-section of migrants living in a number of slums across the developing world today and ask whether those who relocated from rural areas earlier now earn more than those who migrated more recently. They find no relationship between time in the city and earnings in Kenya’s Kibera, and a negative relationship between city tenure and earnings in Bangladesh’s Tongi. If successful people ultimately end up leaving the slum, these facts may re- Intergenerational flect the selection of who remains in the slum over 40 years, upward mobility not a broader lack of upward urban mobility. Yet, it is un- is related to doubtedly true that many of those who come to the city remain poor for decades afterward. urbanisation in A second approach has been to seek cases in which Africa. people, typically immigrants, have been allocated across space by government programmes. Edin et al. (2004) provide a classic example in which the Swedish government directed new immigrants to specific locations across Sweden. However, administrators are rarely willing to completely ignore the idiosyncratic needs of individuals, and so unobserved immigrant characteristics may well have influenced the choice of location and biased the results.6 Sarvimaki et al. (2019) study the forced relocation of 11% of Finns after World War II, in which farmers who were given a similar farm in a different part of the country. Relative to a comparison group who were geographically nearby, the forced migrants were more likely to transition out of farming and into an urban location in the long run and had substantially higher earnings and education over time. Nakamura, Sigurdsson, and Steinsson (2016) study individuals from a wealthy fishing village in Iceland’s Westman Islands whose homes were destroyed by a volcano. Using a spatial discontinuity design, this

6 The so-called Gautreaux Experiment, a US housing desegregation project that placed welfare recipients in randomised locations chosen by the Chicago Housing Authority, is an earlier experiment where apparent administrative randomness was used to estimate the effect of place.

EVIDENCE PAPER / CITIES 10 study shows that 30 years later the displaced workers were more likely to be urban, have a higher education level, and have much higher earnings. A third approach has seen researchers help design social programmes that provide incentives for people to move across space. The US’ Moving to Opportunity for Fair Housing experiment required a randomised share of housing voucher recipients to move to neighbourhoods with lower poverty levels in order to receive the vouchers. Early estimates of the programme found few impacts on the children who moved out of poverty (Katz, Kling, and Liebman 2001). However, more recent work has found quite sizeable impacts on the adult earnings of children who moved out of poverty at an early age (Chetty et al. 2016). Bryan et al. (2014) take a similar approach and provide small incentives (about the cost of one bus fare) for rural Bangladeshi workers to move (at least temporarily) to a nearby city. Even this small incentive generated a 22-percentage-point increase in the number of families reporting that at least one of their family members had sought work in the city, as well as a 33-per- cent increase in average household expenditures. The study also showed that up to three years after the small incentive was paid, treatment households were about 10 percentage points more likely to have a migrant worker in their household. This work suggests real utility gains from moving to the city, because workers continued to come to cities when the incentive was no longer available, perhaps suggesting that initial migration was simply limited by credit constraints. It is worth noting, of course, that small-scale experiments cannot estimate the general equilibrium effects of large-scale migration to the city and may also lack external validity. However, the presence of such drastic effects for those given an opportunity to migrate point to cities’ potential for increasing income and welfare. While the latter studies seem to rule out the possibility that selection fully explains the agglomeration earnings effect, they use data on migrants themselves, and so cannot account for whether urban location generates positive or negative externalities. A final form of experiment shocks the place, not the person, then looks at the impact on the people living in that place beforehand. Greenstone, Hornbeck, and Moretti (2010) measured the differing fates of medium-density communities that did or did not receive the investment generated by the opening of a “Million Dollar Plant” in their area. The results suggest a 12% increase in total factor productivity for incumbent plants, in- dicative of strong positive spillovers that are not internalised in plant opening decisions. This work requires the place-based shock to be independent of unobserved, time-changing attributes at the place level. While the “Million Dollar Plant” experiment comes close, few private or public investment are completely independent of local characteristics. Greenstone et al.’s (2010) results on agglomeration economies open up the possibility that levels of agglomeration are not optimal, but it is not clear that they are directly relevant to developing countries, where movement costs

EVIDENCE PAPER / CITIES 11 may be higher (even in dense areas) and technologies are different. As Glaeser and Gottlieb (2009) emphasise, policy requires comparing the benefits that the winning place gains from having a new plant with the losses that the losing place faces. Designing relocation policies requires us to know the full functional form of agglomeration economies. Imbert et al. (2018) uses variation in international agricultural prices to generate plausibly exogenous variation in earnings across rural areas in China and thus plausibly exogenous variation in the number of migrants moving to nearby urban areas. In-migration leads to a reduction in wages and value added per worker, along with a move to more labour-intensive production. These results seem to suggest a standard downward-sloping demand curve, rather than positive externalities from the in-migration of low skill workers. A firm-level literature links area-level characteristics and plant productivity (Henderson 1999). In this case, the selection problem is that more productive plants may move into more productive places. A parallel “quantities” approach looks at the co-location of industries and tests whether firms locate near other firms that buy and sell their goods, use that same type of workers, or offer an opportunity to exchange ideas, finding evidence for all three -ef fects (Ellison, Glaeser, and Kerr 2010). In the US, co-agglomeration estimates point to the importance of transportation costs for goods and people, at least in manufacturing industries. The literature linking urbanisation with dynamic externalities and national growth is smaller and necessarily less compelling. Many classic theories could also rationalise a causal effect of urbanisation on growth. If fixed- cost technologies required large market sizes, as in Rosentein-Rodan (1943), then urbanisation could provide the “big push” that leads to industrialisation. Cities might enable poorer countries to trade with rich countries. The apparent ease of shopping in Dongguan and Shenzhen’s famous electronics markets for all the parts required to create a state-of-the-art smartphone illustrates this possibility nicely; cities bring together more goods and thus more opportunity in places that might otherwise be lacking. A final hypothesis is that cities enable political change, and dictatorships certainly face more revolutions in more urbanised countries, suggesting that certain governments might have an incentive to slow urbanisation in any way possible. The scale of these theories makes them hard to test. Rauch’s (1993) pi- oneering work estimating human capital spillovers in cities was directly mo- tivated by Lucas (1998)’s paper. In another example, Henderson (2000) links country-level growth and the level of urban primacy and finds a non-monoton- ic relationship. The endogeneity of urbanisation levels and their correlations with other growth-enhancing factors makes causal inference from cross-national data difficult. A more plausible research path may be to examine the links between cities and the ingredients of growth, such as new patent creation and cita- tion (e.g. Jaffe, Trajtenberg, and Henderson, 1993), foreign direct investment

EVIDENCE PAPER / CITIES 12 (Guimaraes, Figueiredo, and Woodward, 2000), and education (Muralidharan and Sundararaman, 2015). Sub-national data makes it easier and more plausible to identify the mechanisms, if any, through which cities are enabling national transitions from poverty to prosperity.

C Can developing-country cities become more productive?

The simple cross-national growth correlation shown in Figure 1b warns that restricting urbanisation may have adverse consequences. Yet, for most developing-country cities, the pressing policy questions are smaller. City governments need to know whether investment in road quality or reforming the permitting process will enhance urban productivity or whether these expenses are unjustified within tight Dense urban city budgets. Transportation infrastructure is one obvious environments— place to look for productivity gains. Firms operate in and the negative particular locations, and they need not only a supply of externalities they physical space, but access to workers, customers, and suppliers. Government involvement in transport infra- give rise to— structure is ubiquitous, because transport infrastructure intensify the need has some of the characteristics of a natural monopoly for government (limited non-rivalry) and usually requires large-scale coordination. As the relationship between transportation, rules that establish building supply, and firm productivity cannot be stud- both the rights and ied through simple partial equilibrium models, we de- obligations of firms. vote Section IV to this topic. Productivity may also benefit from improvement of the legal infrastructure that governs firm behaviour. Dense urban environments—and the negative externalities they give rise to—intensify the need for government rules that establish both the rights and obligations of firms. These rules, if too onerous, can reduce productivity (Djankov et al. 2003), but some regulations that preserve safety and uphold obligations seem likely to be beneficial. Designing the optimal set of rights and obligations is difficult enough under ideal circum- stances, but developing countries often have small budgets and a dearth of effective legal infrastructure (Besley and Burgess 2000, World Justice Project 2019). A system that provides the ability to determine property rights also gives rise to the potential to abuse that power (Goldstein and Udry 2008). It could lead to red tape and inefficiency, causing corruption to spread as a second-best means to fund public goods in the presence of tight government budgets (Banerjee, 1997, Banerjee, Hanna, and Mullainathan 2013). More general-

EVIDENCE PAPER / CITIES 13 ly, the enforcement of any rights or obligations needs some kind of solution to the guardian’s problem (Hurwicz 2008, Björkman and Svensson 2009). On institutions, there is also the question of which sphere of government has responsibility to design and promote urban productivity. Many policies are set at the national level but require strong coordination on the local level to be successful. It is important to understand how local governments can best design and implement policy to leverage their cities’ comparative advantage for job creation and competitive production. Research on institutional improvements requires viable actual or natural experiments, and a small but growing literature now attempts to understand solutions to these problems. Khan, Khwaja, and Olken (2015) work with government employees in Punjab, Pakistan to randomise an incentive pay scheme that rewards property tax collectors for the revenue they raise. They find a large increase in government revenues for little cost in terms of taxpay- er satisfaction or assessment accuracy. In the Kyrgyz Republic, Amodio et al. (2018) provide incentives to reduce bribe-taking among business tax inspec- tors and find that this reduction in bribes is passed through to consumers in the form of lower prices. The work of Banerjee et al. (2014) with the Rajasthan police provides a more nuanced view. The negative results from several seem- ingly sensible strategies serves to remind us of the difficulty of reforming complicated institutions and the importance of nuanced understanding of local systems and politics and natural experimentation in problem solving. The permitting and regulatory environment will be particularly important if local entrepreneurship is a significant determinant of local success, as appears to be true in the US (Glaeser, Kerr, and Kerr 2015). Yet, it is unclear if poor countries need more local entrepreneurship or more foreign direct investment. If developing-country cities today will be built by new versions of Soichiro Honda, the man who began with a small repair shop and grew it into a worldwide automobile empire, then improving permitting and regulatory processes for small businesses is crucial. If foreign inputs are critical, it should lead to an emphasis on making the urban environment more attractive to outside talent and investment. If cities are to Developing country cities, specifically those in Sub- Saharan Africa are the last to benefit from the potential be productive, demographic dividend. It is the only region in the world harnessing the where the youth population is increasing. Yet the young, particularly young women struggle to enter the labour burgeoning young market. The young account for 60% of Africa’s jobless. population in both If cities are to be productive, harnessing the burgeoning formal and informal young population in both formal and informal employment is critical. In addition, this informal economy repre- employment is sents between 50 - 75% of total employment (Chen and critical. Beard, 2018) with similar levels of informal settlements. Understanding how cities can plan for and work with this informality is critical.

EVIDENCE PAPER / CITIES 14 D How can the economic benefits of cities be more widely shared?

Even the most productive cities often contain islands of poverty amidst seas of plenty. Policy makers often want to take action to improve the welfare of their least fortunate citizens. For many urban leaders, the most pressing policy question is how the prosperity of a few can be expanded to include the many. Plato’s Republic famously notes that “Any city, however small, is in fact divided into two, one the city of the poor, the other of the rich”. As successful cities attract both rich and poor people, the existence of urban poverty or inequality is not a sign of urban failure. The important question is whether cities are turning poor people into middle-class people or whether the poor are remaining trapped in perpetual pockets of deprivation. For example, while urban America may be productive, it does not seem to be providing much opportunity for many of its poorer residents. The opportunity atlas of Chetty et al. (2018) documents the low levels of upward mobility for poorer children growing up in America’s cities, finding that the conditions that foster overall labour market productivity are not necessarily the same as the ones that allow for upward mobility. In China, Combes et al. (2019) find that better-educated rural-urban migrants seem to experience much larger wage gains than less-educated workers who come to cities, a phenomenon that Autor (2019) finds echoed in the US. As these studies suggest, individual education is strongly linked with upward mobility in cities (Psacharopoulos and Patrinos, 2018). Schools teach children skills that facilitate communication, such as reading, writing, and grammar, and these skills then enable urban interactions. The overall level of education in a city is also strongly linked to its success, as measured both by earnings (Rauch 1993, Moretti 2004, Chauvin et al. 2017) and population growth (Glaeser, Scheinkman, and Shleifer 1995). Urban density and education appear to be complementary (Glaeser and Resseger 2010), which suggests that better education may enable poorer children to take advantage of urban opportunities. The spatial structure of cities is another way in which inequality is ce- mented, with access to opportunities, social networks and safe and secure housing being concentrated more in some areas rather than others. Understanding these location, network and neighbourhood effects on social mobility and access to job opportunities in the city is also important for ensuring that the gains of agglomeration are more widely shared. One of the main challenges that researchers face in studying how cities provide opportunities for those at the bottom of the distribution is the availability of data. Panel survey data on the poor for example is particularly difficult to collect in cities and even more so in low income countries. To make progress on this agenda, researchers will need to overcome the challenges from

EVIDENCE PAPER / CITIES 15 collecting data on the movement of people into the fringes and margins of cities or urban seasonal migrants. New sources of data, more widely discussed in section 4, could help surmount these challenges.

BOX 1 Next steps and research priorities

Policy makers in developing world cities need evidence on the strength and nature of agglomeration, and the constraints that restrain their productivity potential.

• Measure the size and nature of the returns to concentration across formal and informal activity. — Do cities facilitate matching between firms and workers and encourage the exchange of goods and services? Are they escalators that facilitate the rapid learning of new skills and techniques? — For residents, how do neighbourhoods and slums help or hinder access to economic opportunity and mobility across social and vulnerable classes?

• Understand the forces that reduce these potential agglomeration benefits. — Which constraints on firms —such as a lack of skills, limited access to input and output markets, burdensome regulations or limited energy access—constrain labour demand and support such high levels of unemployment amongst the young and vulnerable population? — What limits developing-country workers’ abilities to acquire skills and learn from employers and co-workers as workers in the developed world? — How do the specific features of developing country cities—unplanned spatial expansion, the persistence of informality across land and labour markets—drag down economic performance?

• How can the comparative advantage of different cities be identified and how can local governments best design and implement policy to leverage this? — What is the role of local government in local economic development policy? — How can national and local government policies for enhancing productivity be better coordinated and complementary? How can national and local government policies for enhancing productivity be better coordinated and complementary?

Evidence on the size and nature of these forces will require a range of empirical work from RCTs and historical policy quasi-experiments to structural modelling, but the answers will help policy makers identify policies with the greatest potential to raise urban productivity.

EVIDENCE PAPER / CITIES 16 3 Infrastructure, institutions, incentives, and density’s downsides

Urban proximity enables the spread of ideas and the sale of services, but it also leads to the quick spread of bacteria and congestion of city roads. In developing countries, urbanisation has proceeded far more quickly than institutional development. Consequently, massive developing-country cities now must face the downsides of density, such as contagious disease, crime, and traffic congestion, with limited wealth and scarce public capacity. In this section, we focus on three central downsides of density: pollution, congestion, and crime. We begin by discussing the costs of different urban harms and then explore several central themes In developing that cut through the policy responses to most urban disamenities, such as behavioural responses to poli- countries, cy changes, social returns to infrastructure and pub- urbanisation lic capacity. We then apply these more general ideas has proceeded to the topics of housing, planning, municipal finance and water and waste management. We end with a far more quickly brief discussion of climate change and sustainability. than institutional It is important to note that while this section development – discusses the downsides of density and the extra service demands created when people live in dense en- creating an urgent vironments, cities are likely to make the provision of need for these services less expensive and more cost-effective com- cities to tackle pared to rural areas. Mass vaccination, high quality primary education, and social protection are exam- the downsides of ples of services that all likely to be easier to deliver in density. urban areas, due in large part to the lower cost of ac- cessing people. Thus, as highlighted above, cities are likely to be productive places, both in the private and public domain. We note that there is comparably less evidence on this dimension, and that estimates of how the cost of service delivery change with density would be important future work.

EVIDENCE PAPER / CITIES 17 A What are the social costs of urban contagion, congestion, and crime?

The first and most basic task is to estimate the size of the costs created by urban disamenities. This question is particularly crucial for policy makers with limited time and resources who are trying to focus on the urban problems that do the most harm. If air pollution does a small amount damage but bad water causes widespread illness, then policy makers may want to invest more in water systems. Conversely, if air pollution is more harmful, then policy makers should turn their regulatory energy toward regulating the corporations and ve- hicles releasing smoke into the air. The economics literature on the impact of urban air pollution is large and compelling. The air pollution literature has focused on the adverse health consequences of bad air quality. Currie, Neidell, and Schmeider (2009) examine air-quality-monitor data in New Jersey and find that infant health suffers as air quality deteriorates. One challenge with this work is that poorer people, who are sicker for many reasons, tend to live in places with worse air. The authors address this issue by looking at air quality changes over time for a panel of families in diverse areas. Alexander and Schwandt (2019) look at air-quality deterioration associated with widespread cheating on automobile inspections and find that bad air increases asthma and decreases birth weight. While these papers focus on the US, there is also a literature, surveyed by Currie and Vogl (2013) that looks at developing-country cities. Arceo- Gomez, Hanna, and Oliva (2016) find that bad air quality has even more serious effects in Mexico City than in the US, suggesting that the health effects of pollution are compounded in poorer places. Cesur, Tekin, and Ulker (2017, 2018) show that switching from coal to natural gas led to air quality improvements in Turkey, which in turn improved health outcomes for children and adults.7 A smaller literature links air pollution to more negative economic outcomes, such as labour supply, and also finds negative effects of air pollution (Hanna and Oliva, 2015, Fan and Grainger, 2019).8 At the city level, air pollution can also harm the local economy by repelling skilled high-productivity individuals (Kahn, 1999). Among economists, Cutler and Miller (2005) and the work of Werner Troesken (e.g., Troesken 2008) has been particularly important in establishing the historic link between water infrastructure and public health. More recently, there has been a dramatic increase in the work of economists on water in the developing world. Gamper-Rabindran, Khan, and Timmins (2010) found that piped water decreased infant mortality in Brazil. Devoto et al. (2012)

7 Quah and Boon (2003) place a dollar value on health outcomes with tools like multiplying mortality estimates by the value of a statistical life. 8 Heath, Mansuri and Rijkers (2018) find that high frequency health shocks significantly reduce female labour supply.

EVIDENCE PAPER / CITIES 18 found that piped water in urban Morocco did not increase health, presumably because families already had access to high-quality non-piped water, but increased happiness by making water procurement easier and less contentious and freeing up more time for leisure. Buchmann et al. (2019) find the particularly striking result that a health campaign to reduce exposure to arsenic-contaminated water increased infant mortality by inducing households in Bangladesh to switch to water sources with higher faecal contamination.9 Traffic congestion is defined by excessive time spent on travel relative to driving on an uncrowded road. Economists have assigned a value to this lost time by multiplying the lost minutes by after-tax wage (Alonso 1964). More sophisticated papers have used survey instruments to find that the cost of time spent in traffic is actually lower than lost wages (Calfee and Winston 1998.10 Investment in transportation infrastructure may lead to either urban growth (Duranton and Turner 2012) or suburbanisation (Baum-Snow 2007). While reduced-form methods can estimate these impacts, interpreting those estimates requires the structural models that we will discuss in Section IV. Most urban leaders accept on faith that reducing crime, and particularly violent crime, to wealthy country norms is desirable. After all, governments have long sought a monopoly on violence. Consequently, the economics of crime and punishment has rarely focused on the costs of crime, but has instead tried to estimate the impact on crime of different policies, such capital punishment (Ehrlich 1975), more policing (Levitt 1997), and lengthier prison sentences (Kessler and Levitt 1999). The most standard approach to estimating the costs of crime is to estimate victims’ losses when crimes do occur (Chalfin 2015), so that murders cost millions and robberies cost hundreds. These costs may overestimate the true social cost of crime, because they omit the benefits of crime to the criminal; however, it seems far more likely that they underestimate the costs, because they neglect the costs of fear and avoidance behaviour for others in the community.11 Ludwig and Cook (2001) use a contingent valuation survey to

9 The economics literature on solid waste management remains as limited as the literature on water before 2000. There is however a sizable epidemiological literature that finds robust correlations between disease and proximity to a wide range of solid waste (Giusti, 2009). 10 While US studies typically assume that traffic speeds in the absence of the congestion would be the speed limit, the poor quality of roads in the developing world can reduce travel speeds considerably, even outside peak hours (Kreindler, 2018). 11 When person A steals person B’s bicycle, then presumably this is a transfer from person B to person A rather than a pure loss of welfare. Applying this logic to murder, however, is more problematic. Even if person A receives some psychic benefit from killing person B, few observers would be willing to include murderous enjoyment as a reasonable element in any social welfare function.

EVIDENCE PAPER / CITIES 19 estimate respondents’ willingness to pay to live in communities without fear of crime. Hedonic price models can also use the difference in housing prices between safe and unsafe areas to estimate the social losses due to fear of crime (Thaler and Rosen 1976). Most estimates find that urban crime, unsur- prisingly, generates significant costs, including spurring on outward migration (Cullen and Levitt 1999) and reducing tourism (Biagi and Detotto 2014). Measuring the average social costs of contagion, congestion and crime remains a research priority. However, the differential impact of the downsides of density also deserves attention. The most vulnerable economically, who live in specific areas, could be more affected by pollution. The poor in slums typically do not have access to reliable water supply and so improving access could have a significant impact on health outcomes.

B Incentives and behavioural change

Most urban infrastructure, such as subways or aqueducts, can be interpreted as adding effective space to a city in which space is scarce. Yet, adding infrastructure may not be as cost effective as reducing the behaviours that require it, especially when added infrastructure induces more socially harmful behaviour. Duranton and Turner (2011) empirically document the “Fundamental Law of Highway Traffic,” which states that overall vehicular miles travelled increase roughly one-for-one with highway miles built. If this law holds, then building more roads does little to solve traffic problems, because the new roads will simply become congested with new drivers. Consequently, the problems associated with density often need some combination of infrastructure and incentives to be effective. The crime and economics literature has long asked whether incentives can reduce harmful behaviour (e.g., Ehrlich 1975, Levitt 1998, Nagin 2013), but much of this US-based literature may not translate easily to developing world Adding infrastructure cities. While over 50% of murders typically lead to an may not be as cost indictment in the US, under 15% of murders in Brazil effective as reducing are solved (Misse and Vargas 2007). Corruption, mal- feasance, and gang power are often worse in develop- the behaviours ing-country cities, so the same incentives may not be that require it, as effective. especially when The pollution and congestion literatures focus more on the impact of regulations than on flexible in- added infrastructure centives. Davis (2008) documents the impact of the induces more Hoyo No Circula programme, which limited cars’ ability to drive on certain days, on air quality in Mexico socially harmful City. Kreindler (2016) similarly shows that license- behaviour.

EVIDENCE PAPER / CITIES 20 plate-based bans on driving effectively reduced con- In congestion, the key gestion in Delhi. The introduction of congestion pricing in behaviour that can London, Stockholm, Oslo, and Singapore all provide reduce the benefits case studies on the impact of pricing roads. Typically, of new infrastructure the best that can be done with these interventions is to compare before and after congestion pricing. It does is driving. In public appear that London’s roads became more passable af- health interventions, ter it imposed its congestion charge (Leape 2006). Yet the usual problem it is not obvious the results for London will generalise to Jakarta, as in Hanna, Kreindler, and Olken (2018). is take-up, where Kreindler (2018) estimates a model of demand people choose not to for driving trips in Bangalore using an experimen- pay connection fees tal structure where individual drivers were randomly offered incentives to avoid peak times on crowded that cover the “last roads. Strikingly, he found that the behavioural ad- mile.” justment was modest, and that Indian roads wouldn’t flow very quickly even if congestion was reduced substantially. This type of experimental model has promise, yet it is essential to bear in mind that any small experiment will change the general equilibrium effects that reach across cities. In congestion, the key behaviour that can reduce the benefits of new infrastructure is driving. In public health interventions, the usual problem is take-up, where people choose not to pay connection fees that cover the “last mile.” Ashraf, Glaeser, and Ponzetto (2016) note that both in New York City historically and African cities today, poorer citizens have often been unwill- ing to pay the marginal cost for connections to cleaner water sources. One empirical question is whether they will connect if given subsidies, or whether the more effective route would be to impose penalties on those who do not connect.

C Estimating the social benefits of infrastructure

Randomised control trial methods are much harder to implement for infrastructure than for incentives, both because infrastructure substantially impacts an entire area and because randomly relocating infrastructure is cost prohib- itive. In some cases, simple difference-in-difference methods can estimate the impact of infrastructure, as Alsan and Goldin (2019) did for sewerage in greater Boston or Duranton and Turner (2011) did for roads within the US. Yet, these estimates may tell us little about a new project in Delhi or Nairobi. The primary tool that economists have brought to infrastructure is cost-benefit analysis, which attempts to catalogue the gains and losses from

EVIDENCE PAPER / CITIES 21 building new roads, tunnels, and sewerage systems. Typically, this work brings together the knowledge of economists and engineers, as in Meyer, Kain, and Wohl (1965). A central result of the early forays into urban infrastructure analysis was that bus systems, sometimes on dedicated lanes, are far more cost effective than rail systems. This analysis helped inspire the Bus Rapid Transit (BRT) systems that have been implemented in Curitiba, Brazil; Bogotá, Colombia; and elsewhere. In early years, the benefits of infrastructure largely focused on the direct benefits to users. Infrastructure Measuring the social boosters would then forecast high projected ridership benefits of public levels, which would then be disputed by economists (Kain 1992). User-fee financing creates some fiscal dis- infrastructure across cipline, since projects are expected to cover their costs, population groups is but if user fees are too low to pay for total or even op- an area that requires erating costs, then that discipline vanishes. Low fees are typically justified because marginal costs are below av- further research – erage costs or because of a desire to redistribute to poor- both to understand er infrastructure users. As infrastructure investment in- for whom and where creasingly relies on alleged agglomeration benefits, the scope for overselling becomes even larger, which only the returns are increases the need for the rigorous structural modelling greatest, but also that we discuss in Section IV. New infrastructure projects are often given prece- the political economy dence over maintenance, which is especially problematic constraints that if there are particularly high returns to maintaining older urban policymakers roads and bridges (Gramlich 1994). This could be particularly true if developing countries are attracting wealthy out- face. side investors who are looking to make an impact and do not do appropriate research on existing infrastructure beforehand. Additionally, the quality of the management of infrastructure will depend on institutions and incentives. While World Bank statistics claim that Lusaka, Zambia, has almost com- plete water connections, in practice, some areas of the city seem to lack viable connections much of the time. Ashraf et al. (2017) show that the reason behind this is that the water company in Lusaka is much quicker to respond to supply problems for customers who pay per litre than customers who pay per month. We turn now to the institutional side of urban management. Here again, measuring the social benefits of public infrastructure across population groups is an area that requires further research. This is not only to measure for whom and where the returns to public infrastructure could be the greatest but also to shed light on the political economy constraints that urban policymakers may face when deciding on where the next water pipe should be built. We turn now to the institutional side of urban management.

EVIDENCE PAPER / CITIES 22 D Institutional reform and public capacity

Public institutional capacity is a precondition for any meaningful reform. However, it is often difficult to use modern empirical methods, such as randomised control trials, to understand paths toward better institutions. Some studies measure whether changes in incentives can alter the behaviour of public officials. Muralidharan and Sundararaman (2011) show that Indian teach- ers appear at work more often when pay is linked to performance. Ferraz and Finan (2011) show that federal auditing of mayors in Brazil reduces corruption. Yet the impact of any incentive programme can be easily distorted in a corrupt institution; Corruption could lead to misuse of incentives, information about upcoming audits in exchange for pay, etc. In many developing countries, proving that an innovation can work is not the same as showing that it will actually change institutional performance. Most work on the institutions that matter for developing-country cities is descriptive or involves simple comparison. For example, Engel, Fischer, and Galetovic (2014) present a magisterial overview of public-private partnerships (PPPs) throughout the world. Their work typically reviews the performance of PPPs and often com- Economic theory pares their performance with governmental alterna- makes predictions tives. Singh (2018) presents a similar study comparing the roughness of Indian roads that are maintained by about the impact public and private entities. In this case, the private of limitations on roads are far smoother than their public alternatives. While Singh’s study is persuasive, in general, property rights, the private provision of public services has a far more but there is little mixed track record. As Engel, Fischer, and Galetovic research that fully (2014) show, private companies often manage to rene- gotiate with public entities to radically increase their teases apart the compensation, leading to a potentially higher cost for impact of partial the government than if they had supplied the need control over urban themselves. Theoretically, private entities should have better incentives to maintain quality because they can land. only reap returns if customers use them, but in some cases, quality is average or even poor. Certainly, private entities that are paid with public money have a strong incentive to subvert the system and extract more public resources wherever possible. While much institutional research focuses on the executive branch, the judiciary is also critical, for every market failure is ultimately a failure in the maintenance of property rights (Coase 1960). If courts fail to protect land rights, then people lack the incentive to develop that land. When courts fail, ordinary people waste time protecting their property from expropriation (Field 2007).

EVIDENCE PAPER / CITIES 23 Property rights over urban land are actually a nexus of rights including the rights to use, develop, sell, rent, and mortgage. In many developing-country cities, these rights are far more fragmented than they are in the West. For example, the residents of informal settlements may well be protected in their right to use a piece of land, but since they have no title, they cannot sell that land or mortgage their property to start a business (de Soto, 2000). Economic theory makes predictions about the impact of limitations on property rights, but there is little research that fully teases apart the impact of partial control over urban land. More broadly, many cities in developing countries are expanding their urban land cover, rather than building higher levels of density. According to Angel (2011) if density-levels remains the same in African cities, urban land cover in 2050 will be four times higher than it was in 2000. This is compounded by institutional structures unable to keep up with the expanded economic footprint of fast-growing cities: often, many government units are responsible for urban service delivery in a single city, with little coordination between them (Collier et al., forthcoming). There is little research from developing countries that provides economic analyses of the impact of The lack of public spatially fragmented governance structures on cities. funds represents Particularly, along two lines: first, the impact on the a vicious cycle: a productivity-levels in a given city. While there is some evidence of how fragmentation reduces urban produc- city without public tivity-levels from OECD countries, there is none from capacity may find it developing countries (OECD, 2015). Second, on the difficult to collect the impact of spatial fragmentation on equity concerns, particularly in cases where fragmentation coexists with tax that it needs to fiscal decentralisation. Intuitively, it would be harder fix its public capacity in administratively fragmented cities for local govern- problems. ments to redistributing revenue between wealthy and poorer regions. This is particularly of concern due to stagnation of the urban cores in many developing cities, for example, in almost all mega cities in South Asia (Peter and Mark, 2016).

E Housing, planning, public finance, and water and waste management

Four large policy areas relate to the downsides of density and the issues we have just discussed: local finance, water and waste management, housing, and urban planning and zoning. We now briefly discuss the ways in which public capacity, behavioural responses, and cost-benefit analysis play out in these core areas of urban policy making in the developing world.

EVIDENCE PAPER / CITIES 24 Municipal finance is particularly central to almost all efforts to improve urban quality of life. If a city can’t raise public funds, then it will have trouble providing better police, fixing potholes, or managing waste. In addition, this problem represents a vicious cycle: a city without public capacity may find it difficult to collect the tax that it needs to fix its public capacity problems. Consequently, cities can get caught in a low capacity/low revenue trap that may be particularly problematic. It is thus valuable to learn whether certain taxes, such as simple land value taxes, are easier to collect than others in weak-capacity environments. Naturally, the ease of collection needs to be weighed against the other behavioural distortions induced by specific tax rates. There is a large literature on the behavioural impact of different tax rates in the wealthy world. The literature is smaller in the developing world, and there is a need to understand which specific taxes might deter workers and firms from entering into the formal sector. The long-run research goal should be to generate serious cost-benefit analysis of different taxes by combining evidence on implementation challenges with evidence on the magnitude of distortions. When cities fail to provide decent waste and Today, households water management, their residents face the ancient urban scourge of contagious disease. The externali- appear to have a ties that come from disease and waste disposal ex- greater willingness plain why governments have been engaged with these to pay for better issues at least since Rome built its Cloaca Maxima, one of the world’s first sewage systems, under the water than to pay Tarquinian kings. Today, households appear to have to dispose their a greater willingness to pay for better water than to waste, because it pay to dispose their waste, because it is their neighbours who mostly pay for accumulated rubbish, but is their neighbours more research estimating private willingness to pay who mostly pay for water and waste management is important. Such work needs to be combined with larger estimates of for accumulated the costs of water- and waste-management failure to rubbish. understand the total benefits of providing better services. The gap between total benefit and private benefit can help inform any public decision to either subsidise adoption of better services or tax non-adoption of those services. Since water and waste management typically involve significant infrastructure, these policy areas also involve institutional choices. When should these services be provided be governments, and when should cities turn to private provision? How does the institutional nature of service provision determine service quality and access? Are there particular local characteristics, such as public capacity, that should shape the choice of institutional reforms? Cities typically manage their physical land areas, both through land- use planning and housing policy. These policy actions can have profound

EVIDENCE PAPER / CITIES 25 long-run impacts on housing costs, urban mobility, and the very shape of the city. Once again, policy action should be informed by basic studies that estimate the far-ranging impacts of planning decisions. The decision to provide public funding for housing can become better informed if there are better estimates of the long-run impacts of such housing on economic and social outcomes for its residents. In addition, the decision of where to place this housing is of critical importance, given that housing is more than simply shelter – but rather a proxy for many other urban ‘goods’, such as connectivity and social networks. Assessing the benefits of concepts such as mixed-income housing, as well as fully understanding the costs and benefits of locating social or public housing on high value land in the urban centre would be very useful in informing future human settlement policies. The treatment of informal housing (or ‘slums’) is also an important area to understand. Formal housing comes bundled with a series of obligations aimed at overcoming the externalities of dense living. Space is provided for transport access, sanitation and water, and building regulations ensure that low quality construction does not threaten neighbours assets. These provisions, how- The decision of ever, are costly and may limit the ability of the poor- where to place this est of the poor, and recent rural to urban migrants to housing is of critical reap the benefits of the city’s density. The treatment of slum areas requires careful weighing of these costs and importance, given benefits. We discuss the issues and evidence through that housing is more the lens of specific structural models in section 3 be- than simply shelter low, but note here that general equilibrium impacts of projects in slum areas make them hard to evaluate – but rather a proxy and may also lead to misinterpretation. For example, for many other urban a slum improvement programme may lead slum dwell- ‘goods’, such as ers to sell off their newly improved homes and create a new slum. If observers concentrate on the goal of re- connectivity and moving slums, this could be interpreted as a failure. social networks. However, this interpretation ignores the fact the original slum dweller has had an increase in wealth, and the household that has moved in has also likely gained, moving from a less desirable location. Comprehensive studies that capture these complex equilibrium effects are important in formulating appropriate policy. Access to public housing can be evaluated through randomised control trials, but changes to zoning rules are far more complex to evaluate. As local land-use rules can have complicated impacts that reverberate through the city, the structural models we will discuss in the next section seem particularly well suited for evaluating those rules. A final important question concerns the institutional choices for housing and planning authorities: How can public entities acquire the capacity to do these tasks well?

EVIDENCE PAPER / CITIES 26 F Cities and climate change

We end this section by noting the particularly critical issue of climate change, which may end up generating large costs for many of the world’s cities. Holding income constant, urban density is associated with lower, not higher, carbon emissions (Glaeser and Kahn, 2010). This is so because proximity fa- cilitates the use of public transport, it reduces commuting distances and smaller homes in urban areas usually come with lower energy use. Moreover, many of the risks associated with climate change are far larger for subsistence farmers than for urbanites who are enmeshed in a global trading system, where food can be provided by formerly colder areas that may become more productive due to climate change. This stresses the need to promote urbanisation not just for its direct impact on structural transformation and growth but also to ensure that growth is sustainable, by reducing the magnitude of global externalities, and to protect the most vulnerable in rural areas for the detrimental effects of climate change. While conditional on wealth, urbanisation reduces emissions, because cities are responsible for about 75% of global carbon emissions, and as growth and structural transformation in developing countries will push this number up- Cities are responsible ward, there is a significant need for research to think for about 75% of about how cities can become more efficient and com- global carbon pact. One need in particular is to think about local externalities from energy consumption. Most of the emissions, and as world’s most polluted cities are now in low and mid- growth and structural dle income countries (see Figure 8 in the Energy and transformation in Environment IGC Evidence paper). The social cost of pollution and other local and global externalities in developing countries these cities is now so large that it risks reducing urban will push this number density going forward. For example, urban density in upward, there is a China has decreased by an estimated 25% in the past few years.12 Many of the world’s cities by 2050 are yet significant need for to be built, especially in Africa, and this offers an op- research to think portunity to think about how cities can be designed in about how cities can a way that reduces both global and local externalities. Research is particularly needed to measure the become more efficient effect of urban policies aimed at reducing the magni- and compact. tude of carbon emissions. These include, for example, the impact of bus rapid transit systems, the effect of tax policies to incentivise the adoption of more energy-efficient residential buildings or, more broadly, to induce more energy-efficient practices.

12 Source : Pew Research Center, available at: https://www.pewresearch.org/fact- tank/2015/11/19/building-outpaces-population-growth-in-many-of-chinas-urban-areas/

EVIDENCE PAPER / CITIES 27 On this issue of supporting the policy effort to mitigate global externalities from urban areas, there is probably greater potential for work on cities of in- termediate sizes. The majority of the world’s urbanites live in cities of less than 500,000 inhabitants.13 Strategies to reduce local pollution deserve more attention as a growth issue not just because of their direct effect on health, but also because effective policies to reduce pollution may also prove important for climate change and hence for generating sustainable growth. The simpler political economy of local pollution policies – where the short-term costs are counter-balanced by short-term benefits – means that such policies are likely to have immediate spillover benefits for climate change, as emissions drop, as well as longer term benefits, by paving the way for broader climate change policies. Many other policies and regulatory reforms to reduce carbon emissions at the city level should be explored. The expected effects of climate change, especially on coastal areas require that we also think carefully about adaptation and in particular the location of cities. Kahn (2010) argues that poorer countries will be able to adapt to climate change by moving population centres inland and towards higher el- evation areas. As long as sea levels rise slowly, the adaptation process that Kahn envisions may be plausible. Strategies to But if climate change is related to rare natural disas- reduce local ters, such as cyclones and tidal waves, then cities – par- pollution deserve ticularly those in coastal areas – face tremendous risk. In designing policies to mitigate and adapt to more attention as the effects of these growth-generated externalities, a growth issue not the question of where authority lies between the cen- just because of tral government and local municipalities is of critical importance. Addressing externalities effectively their direct effect and efficiently often requires innovation at both the on health, but also local and national level. There is a thus need to design governance structures that facilitate these poli- because effective cy innovations, while keeping a certain level of cen- policies to reduce tral oversight and control to minimise distortions pollution may also across cities. Finally, political constraints on policy in this area are often stringent; thinking about social prove important norms about environment quality is also important for climate change when thinking about the design of urban policies for and generating climate change. sustainable growth.

13 Source: U, World Urbanization prospects

EVIDENCE PAPER / CITIES 28 BOX 2 Next steps and research priorities

• How can sanitation and health services can be provided to effectively maintain public health and a clean environment? More generally, how do the social costs of urban density and social benefits of public services vary across population groups?

• Can governments facilitate functioning land markets to allocate space efficiently and provide affordable housing for lower-income residents? — Measure the impact of public housing projects, slum upgrading programmes, and land readjustment, on resident welfare, land prices, productive activity and fiscal costs. — Measure the impact of location of public housing, as well as spatial integration of lower and higher income residents on resident welfare, land prices, productive activity and fiscal costs. — Find evidence on the effect of land use regulations, including rules that promote or reduce economic inclusion.

• How can developing countries improve urban mobility? — Measuring the costs of congestion, which include not only hours lost to traffic but also distortions in the land and labour markets — Evidence on the role of informal networks that dominate transit in African cities. Can these networks complement the more expensive mass rapid transit being introduced across the developing world? Is it more cost-effective to invest in improving those networks or to introduce traditional mass transit systems? What do new technologies like ride sharing pose for the future of mobility in these cities?

• How can developing countries manage and finance service provision and the functioning of local government? — Are there ways to enhance existing tax design, enforcement and compliance at the local level?

— What can new instruments, such as programmes that capture land value, do for areas with low state capacity and high rates of informality?

• How can developing countries build more efficient, more compact cities? — How large are the benefits from more compact cities, and what policies can incentivise this? — How should cities increase public transit adoption, and what is the impact on emissions? — What policies incentivise the adoption of more energy-efficient practices?

EVIDENCE PAPER / CITIES 29 4 The structural approach to transportation and land use

To many architects and land-use planners, the city is synonymous with the built environment. While urban economics emphasises that cities are better seen as massed humanity, the physical city is still profoundly important. Land- use planning plays a particularly central role for many city governments. Yet, economists have typically had Changes to little to say about efficient land-use rules or the costs of bad planning. The growing field of formal spatial large-scale urban modelling offers the possibility of delivering pragmatic investments are tools that can help policy makers plan better and more more akin to changes fully anticipate the far-ranging impacts of any large- scale change to the built environment. in macroeconomic The randomised control trial approach to de- policies, such velopment economics is ideal when considering target- as fiscal policy, ed interventions that are akin to medical drug trials. However, changes to large-scale urban investments are which reverberate more akin to changes in macroeconomic policies, such throughout the layers as fiscal policy, which reverberate throughout the lay- of the economy. ers of the economy. Just as macroeconomics has turned to simulations using tools like dynamic stochastic general equilibrium models, urban economics has begun using complex structural models that largely rely on simulations to understand how new investments or policies will change life within a city.

A The basic form of structural urban models The first wave of urban models made drastic simplifications that reduce cities to a sequence of locations that differ only in their distance to a central business district (Alonso 1964, Mills 1967, Muth 1969). A day spent exploring a real city’s streets shows how this belies the immensely rich spatial differences that make cities so complex and interesting. Realistically, economic activity occurs in locations that vary in terms of air quality, crime rate, infrastructure, and access to shops and restaurants. Recent models have combined rich spatially dis- aggregated data with tools from the trade and economic geography literature to address this richness (see Redding and Rossi-Hansberg (2016) for a comprehensive review). These frameworks allow researchers to quantify the aggregate implications of economic policies, assess how their impacts reverberate through agents’ behavioural responses and linkages across space, and simu-

EVIDENCE PAPER / CITIES 30 late the effect of counterfactual policies to evaluate how competing approach- es might best achieve policy goals. Quantitative models consist of a series of building blocks whose elements are chosen to fit the focus of the research question and the type of data available: geography, workers, firms, the supply of land and housing, and general equilibrium conditions. The geography of a city is comprised of a large number of discrete locations (such as census tracts or blocks). They differ in attributes such as the time it takes to commute to every other location, the supply of land available, and other exogenous characteristics (such as views, access to roads, or the type and slope of land) that affect the city’s available amenities, productivity levels, or the costs of housing development. Workers must choose where to live and work across pairs of locations. This choice depends on attributes that determine how attractive locations are to live in (e.g., their level of amenities and residential floor space prices), work in (e.g., the wage paid by firms), as well as on the cost of commuting between each pair. Depending on the model, residents can differ in their attributes (such as education or location of prior residence, as in Tsivanidis 2019 or Bryan and Morten 2019), make additional choices such as where to shop or which mode of transit to commute with (Allen, Arkolakis, and Li 2015), or have idiosyncratic preferences for each live-work pair (generating upward sloping resident and labour supply curves as functions of location attributes, as in Ahlfeldt et al. 2015). Similarly, firms must choose their locations. Production can potentially take place in any location, but depends on characteristics like productivity, access to labour, and supply of commercial floor space. Technologies can allow for perfect or imperfect competition, constant or increasing returns, fixed or free entry (Redding 2016), multiple industries (Caliendo, Dvorkin, and Parro 2019), and differing extents of firm mobility (Fajgelbaum et al. 2018). Housing supply and usable production space is constructed by developers using capital and developable land available in each location. Land use is determined by landowners who trade off the return to residential or commercial use, potentially subject to zoning restrictions (Ahlfeldt et al. 2015). These individual optimisation decisions are then connected through general equilibrium market clearing conditions that equate the demand and supply for each factor in each location and pin down prices. For example, equating the demand and supply for labour and floorspace determine wages and house prices respectively. These models also allow for externalities: the levels of productivity, amenities, or travel time across (pairs of) locations are often endogenous (Ahlfeldt et al. 2015, Fajgelbaum and Schaal 2019). In this case, the levels of these variables taken as given by agents must be consistent with equilibrium choices. Once the researcher fully specifies the model, three steps must be taken in order to conduct quantitative analysis. First, it is necessary to estimate the

EVIDENCE PAPER / CITIES 31 “deep” parameters assumed to be invariant to the counterfactual policy. These typically consist of elasticities that govern, for example, the sensitivity of commute choices to commute costs or housing supply to housing prices. Second, the model’s unobserved location characteristics (such as amenities and productivity levels) must be recovered. These models are typically exactly identified, so that there exists a unique mapping from observed data (such as residence, employment, and house prices in each location) to unobservables given the model’s deep parameters. Third, the researcher can use the now-identified system of equilibrium equations to simulate the effects of alternative policy sce- narios (such as new transport infrastructure or zoning regulations) on the full urban equilibrium.

B What’s different in developing countries? The majority of these models have been developed within the contexts of cities in rich countries. Should we expect the results of frameworks built to fit Chicago or Berlin to translate seamlessly to Mumbai or Nairobi? Transit and land use are vastly different in cities of the emerging world, characterised by poverty, informality, and coordination problems. The options available to fi- nancially and capacity-constrained governments also differ. We now discuss recent work that has sought to adapt quantitative models to the context of cities in the developing world and outline areas of promise for future work. Bus rapid transit (BRT) systems have quickly become a popular alternative to subways in developing-country cities. They provide similar reduc- tions in commute times at a fraction of the construction cost. New public transit such as BRT may also have profound distributional implications, since the poor rely on public transit that is often slow in these settings due to the over- supply and lack of coordination among informal minibuses. In his analysis of the world’s largest system in Bogotá, Colombia, Tsivanidis (2019) develops a model that allows for multiple skill groups of workers with non-homothetic preferences over different modes of transit. By accounting for the impacts of transit on the residence, employment, and transit mode choices of heteroge- neous workers, Tsivanidis uses the model to trace out the system’s impact on aggregate performance—not only through reducing time lost in transit, but also by improving the allocation between workers, places of employment, and places of residence. He finds that welfare gains are 20-40% larger after accounting for reallocation and general equilibrium effects. Quantitative models can provide insights into what other policies might complement expensive infrastructure to maximise returns on investments. Tsivanidis shows the feeder bus system that reduces the last-mile problem of getting residents from poor, dense neighbourhoods at the city’s edge to the BRT improve welfare more than any single trunk line. He also runs a counterfactual exercise to show that welfare gains would have been 18% larger

EVIDENCE PAPER / CITIES 32 had the government implemented a land value capture scheme in which zoning densities were increased near stations with permits to build auctioned off to developers. Revenues from the permit sales would have covered around 10- 40% of construction costs depending on the extent of in-migration from the rest of Colombia.14 Future work in developing-country cities needs to incorporate more features of transit. However, we need to meet a few requirements to be able to develop smart policy. First, we need evidence that quantifies the wider costs of congestion given their potential distortion of the behaviour of firms and residents. For example, we need to consider that new infrastructure may have different effects in Nairobi or Lagos than in Berlin or Bogotá due to the vast informality of jobs and housing. Second, these models need to confront the fact that most of the public transit in developing-country cities is informal. Tools from industrial organisation combined with recent work on routing and congestion (Allen and Arkolakis 2019, Fajgelbaum and Schaal 2019) should be used to understand how this industry responds to mass transit interventions, how policy makers can ensure their policy complements rather than competes with it, and what other forms of regulation could improve its performance. Third, new technologies such as ridesharing are changing the nature of mobility in cities. Work is needed to understand how developing country-specific variants, such as motorbike hailing in Bangkok or the Uber bus service currently being piloted in Cairo, will impact mobility, demand for cars, and existing public transit services. Land markets in developing-country cities are characterised by a high rate of informality. To understand patterns of land use and density in these contexts, Henderson, Regan, and Venables (2016) develop a structural, dynamic, monocentric city model that allow for formal and informal construction. They use the estimated model to infer high costs of converting slums to formal housing. Gechter and Tsivanidis (2018) develop a quantitative model that allows for formal and informal housing. They use the framework to quantify the implications for equity and efficiency of the redevelopment of Mumbai’s 58 textile mills during the 2000s. They find that the redevelopment increased the stock of formal housing in the city centre, but also that it displaced poor residents from nearby slums whose homes were converted following the ensuing housing price appreciation. Policy makers in developing-country cities need to take informal housing into account when making decisions; ignoring it can have disastrous consequences for citizens and the economy. The presence of externalities in cities suggests the potential for efficiency gains from zoning and land use policies. But more evidence is needed on the size and nature of these spillovers. For example, if businesses become more productive when they cluster together, governments may want to zone

14 See Suzuki et al. (2015) for a comprehensive review of land value capture instruments.

EVIDENCE PAPER / CITIES 33 centrally located land to commercial use in order to Policy makers in promote socially beneficial levels of concentration. However excess traffic congestion or pollution from developing-country certain industries would make such a configuration cities need to take less desirable. Once more empirical work has identified these externalities, since these policies are difficult informal housing to randomise there is a role for quantitative models to into account when use these estimates to inform policy makers about the making decisions; consequences of alternative zoning or land use policies. For example, Allen, Arkolakis, and Li (2015) develop ignoring it can a model that allows them to characterise optimal zon- have disastrous ing in Chicago around an observed equilibrium, while consequences for Bird and Venables (2019) apply a similar framework to evaluate the impact of tenure reform in Kampala. citizens and the The prevalence of rent control, density restric- economy. tions, mixed-use zoning, and minimum floor space requirements for formal housing sector construction in developing-country cities suggests a need for more work in this area. Governments will also spend vast sums on housing investments that reshape the structure of cities, from slum upgrading (Harari and Wong 2018) to con- structing massive new housing developments at the urban periphery (Franklin 2019). Quantitative work should strive to understand the trade-offs, equilibrium implications, and unintended consequences associated with this menu of options. The degree of shared prosperity that arises from transit and housing policy also depends on the sorting response of residents. Will new transit or slum developments that increase surrounding property prices simply benefit rich landowners and displace poor renters in the long run? Tsivanidis (2019) shows that Bogotá’s BRT increased the spatial segregation between low- and high- skilled workers, a feature that is replicated by the model due to the non-homothetic preferences for residential amenities. Couture et al. (2019) develop a model with non-homotheticities and find that sorting responses and endogenous amenities amplified the increase in wealth inequality in the US since 1990 by 1.7 percentage points in terms of welfare inequality. More work to understand the sorting of residents in developing-country cities and its implications for the distributional consequences of spatial policies is clearly needed. Finally, these models should address the coordination problems particularly salient in land markets of the developing world, where urban growth is typically haphazard, unorganised, and sprawling. Exploring the ring of vacant land surrounding Detroit’s central business district, Owens, Rossi- Hansberg, and Sarte (2019) highlight the coordination problems between residents and developers in the presence of residential externalities. When amenities depend on the number of residents, land may remain vacant even if its fundamentals are sound. Dynamic inefficiencies may arise, for example, if land

EVIDENCE PAPER / CITIES 34 use is sticky and agents fail to internalise agglomeration externalities in production. As more migrants arrive in a city, it may simply run out of plots large enough to allow for a concentration of large manufacturing plants in accessi- ble areas.15 Empirical work by Brandily and Rauch (2018) and Michaels et al. (2018) highlight the dynamic consequences of land-use planning in African cities. The dynamic quantitative models of Desmet and Rossi-Hansberg (2015) and Caliendo, Dvorkin, and Parro (2019) could be extended to understand these effects.

C Providing better parameter estimates to make structural models more useful If quantitative models are to provide useful policy insights, their results need to be trusted. First, researchers must establish that their model captures relevant features of the data or (ideally) can replicate the real-world response to a policy change. Second, they must provide credible estimates of the model’s “deep,” policy-invariant parameters. The increasing availability of new, large-scale These models sources of data in developing country cities provides should address an immense potential to validate and estimate these the coordination models in the contexts of quasi-natural experiments or, occasionally, through randomised interventions. problems The most basic form of model validation in- particularly salient volves showing that key parametric relationships in land markets defined in the model capture the data features relevant to the question at hand. For example, if a mod- of the developing el is used to simulate the impact of new transit in- world, where frastructure, then the relationship between commute urban growth is times and behaviour posited by the model should provide a tight fit to the data. Ahlfeldt et al. (2015) and typically haphazard, Monte, Redding, and Rossi-Hansberg (2018) show unorganised, and how the log-linear gravity equations for commuting and migration delivered by their models fit the data in sprawling. Germany and the United States respectively. Our trust in these models increases if they can replicate the response of cities to real-world policy changes. Heblich, Redding, and Sturm (2019) estimate a quantitative model using one year of data from historical London, and then feed in a sequence of new commute times induced by the expansion of the city’s railway system over an eighty-year period. They find the model can

15 Gollin, Jedwab, and Vollrath (2016) discuss the service-led nature of urbanisation in African cities, which have missed out on the higher rates of industrialisation commensurate with urban growth in other continents.

EVIDENCE PAPER / CITIES 35 replicate the gradual concentration of employment in the city centre despite not being targeted in estimation. Tsivanidis (2019) shows that in a wide class of gravity models, the impact of changing transit infrastructure on equilibrium outcomes, such as population or house prices, is summarised solely by its effect on model-defined measures of accessibility. These models predict these relationships to be log linear. Using the variation in accessibility provided by the construction of Bogotá’s BRT, he shows this is precisely what occurs in the data. Future work should leverage the increasingly available high-frequency data discussed below to incorporate pre-analysis plans into structural work. If researchers can show that quantitative models accurately predict the effects of new infrastructure or other policy interventions they have yet to see, these models’ insights will become yet more believable. The next task is to credibly estimate a model’s parameters. Some randomised interventions do exist. Akram, Chowdhury, and Mobarak (2018) assess the equilibrium impacts of urban emigration on rural villages by randomly varying the fraction of residents offered transport subsidies. Brooks and Donovan (2019) randomly construct bridges across Nicaraguan villages to evaluate their effects on reducing the market-access risks posed by seasonal flash floods. In a more urban context, Gonzalez-Navarro and Quintana- Domeque (2016) exploit randomisation in road upgrades across Mexican neighbourhoods to examine their impact on housing prices. A second approach is to estimate the parameters of a structural model by matching reduced form coefficients from (quasi-)experimental settings. Fogli and Guerrieri (2019) examine the extent to which spatial sorting and neighbourhood effects amplify wealth inequality. The authors estimate the parameter governing the strength of neighbourhood effects by ensuring a one-standard-deviation increase in neighbourhood quality, as a child delivers a 10% higher income than their parents in their model simulations, precisely the estimate from Chetty and Hendren (2018).16 Randomised housing interventions in developing-country cities, such as the Ethiopian public housing lot- tery studied by Franklin (2019), could provide new sets of relevant estimates to calibrate these models. The third and most common approach is to use quasi-natural experiments directly as sources of identifying variation. This has long been popular in trade and economic geography (Donaldson and Hornbeck, 2016; Donaldson, 2018), but has also recently become increasingly popular in urban economics. The seminal work by Ahlfeldt et al. (2015) exploits the construction and fall of the Berlin wall as quasi-random variation in the density of economic activity. This allows them to estimate the strength of agglomeration spillovers

16 Faber and Gaubert (2018) estimate the spillover parameters of a quantitative spatial model in Mexico through an indirect inference approach. They ensure that the coefficient from an IV regression of employment on tourism attractiveness using data generated from their model matches that of the reduced-form analysis.

EVIDENCE PAPER / CITIES 36 across space. Recent examples in Colombia and India use large-scale transit and land-use policy changes to estimate quantitative urban models in poorer countries (Tsivanidis and Gechter 2018, Tsivanidis 2019). Quantitative work has so far focused on rich countries due to data availability, but new sources of large-scale data will allow researchers to increasingly apply this class of models to cities of the developing world. Machine vision techniques have opened the possibility of using daytime satellite imagery to measure the size and density of slums (Gechter and Tsivanidis 2019) and urban areas (Vogel et al. 2019). Google Streetview can be used to predict income by measuring the attractiveness of neighbourhoods (Naik et al. 2015). Cell-phone metadata, Google Maps, and credit card data can be used to measure commute flows, congestion, and consumption across space (Blondel, Decuyper, and Krings 2015, Kreindler and Miyauchi 2019, Akbar et al. 2019, Donaldson et al. 2019). Quantitative work Of course, these datasets have drawbacks, often related to sample selection. The population who has so far focused use cell phones and credit cards may be very differ- on rich countries due ent than the population who do not. This threatens to to data availability, bias analyses and runs the risk of steering urban work toward higher-income settings where digital traces are but new sources of available. While large-scale administrative datasets large-scale data will are promising (albeit with their own concerns of mis- allow researchers to reporting), there remains an important value in primary collection efforts to uncover high-quality, repre- increasingly apply sentative data to complement and validate those from this class of models alternative sources. to cities of the Structural work has limitations. These models make strong functional form assumptions for tracta- developing world. bility that are typically log-linear. Parameter estimates will therefore reflect first-order approximations around an observed equilibrium but may no longer be invariant to large policy changes considered in coun- terfactuals. Slight deviations from these functional forms may deliver very different policy implications (Glaeser and Gottlieb 2008). Static models used to evaluate the impact of transit infrastructure changes, for example, may ignore the adjustment costs involved when individuals need to relocate from one neighbourhood to another, or the larger impacts this churn may have on their children. The results of structural models should therefore be considered an additional input for informing policy by quantifying the effects of alternative options along clearly stated dimensions, rather than a sole guide to policy decisions.

EVIDENCE PAPER / CITIES 37 BOX 3 Next steps and research priorities

Two classes of research questions are particularly well-suited to the structural approach.

• Measuring spillovers, linkages and general equilibrium effects — For example, investments in infrastructure, housing stock or public goods like education induce a host of sorting responses by households and firms as well as feedback effects through prices.

• Understanding the aggregate impacts of recent policies or potential impacts of future reforms. — How large are the aggregate gains and tax revenues from infrastructure investments? — Given the unplanned expansion of cities, is the current spatial configuration of the city (such as the location of ports, markets and schools in central areas) efficient given the current urban organisation and the opportunity cost of allocating that land to other purposes? What zoning or land use planning policies could improve the current and future spatial configuration of cities, given that so many of the world’s cities have yet to be built?

EVIDENCE PAPER / CITIES 38 5 Conclusion

The population of the world’s poorest cities is expanding massively and will continue to do so over the coming years. The existing empirical evidence suggests that agglomeration economies may be particularly large in the developing world, implying that urbanisation can provide a pathway from poverty to prosperity. Large cities in Africa and South Asia have long been conduits for economic exchange between poor countries and rich countries, enabling trade and the spread of knowledge, crucial ingredients for long-term growth. Yet, even when rural people migrate to cities to take advantage of this greater level of economic opportunity, many of them remain poor, often rel- egated to living in slums for decades (Marx, Stoker, and Suri 2013). Slum dwellers face risks from criminal gangs and contagious disease. Even beyond these dangers, many urbanites struggle with long commutes and relatively high housing costs. More effective government policy may be able to allevi- ate these downsides of urbanisation, and more research is critical to learning how to make government more effective. A hope is that the process of urbanisation itself will lead to The cities of the improvements in governmental accountability and competence. developing world are We conclude this paper with one clear mes- the stage on which sage: The cities of the developing world are the stage the lives of billions on which the lives of billions of people will be played out. These places are vitally important to the future of of people will be the world and deserve far more research. Two thirds played out. These of Africa’s cities are yet to be built; if these cities can places are vitally be made safer and more efficient, then the prospects for Africa’s economic growth could be enormous. It is important to the only by designing smart, data-driven policy for cities future of the world that developing countries can make significant pro- and deserve far gress toward economic and social well-being in the years to come. more research.

EVIDENCE PAPER / CITIES 39 6 References

Ahlfeldt, Gabriel M., and Elisabetta Pietrostefani. Alonso, William. Location and Land Use: Toward a “The Economic Effects of Density: A Synthesis.” General Theory of Land Rent. Erscheinungsort nicht Journal of Urban Economics 111 (May 1, 2019): ermittelbar: Harvard University Press, 1964. https:// 93–107. https://doi.org/10.1016/j.jue.2019.04.006. www.degruyter.com/document/doi/10.4159/harvard.9780674730854/html. Ahlfeldt, Gabriel M., Stephen J. Redding, Daniel M. Sturm, and Nikolaus Wolf. “The Economics Alsan, Marcella, and Claudia Goldin. “Water- of Density: Evidence From the Berlin Wall.” Econo- sheds in Child Mortality: The Role of Effective Water metrica 83, no. 6 (2015): 2127–89. https://doi. and Sewerage Infrastructure, 1880–1920.” Journal org/10.3982/ECTA10876. of Political Economy 127, no. 2 (2019): 586–638. https://doi.org/10.1086/700766. Akbar, Prottoy A, Victor Couture, Gilles Duranton, and Adam Storeygard. “Mobility and Congestion Amodio, Francesco, Jieun Choi, Giacomo De in Urban India.” Working Paper. National Bureau of Giorgi, and Aminur Rahman. “Bribes vs. Taxes: Economic Research, November 2018. https://doi. Market Structure and Incentives.” Working Paper. org/10.3386/w25218. IZA Discussion Papers. Institute of Labor Economics (IZA), July 2018. https://ideas.repec.org/p/iza/iza- Akram, Agha Ali, Shyamal Chowdhury, and dps/dp11668.html. Ahmed Mushfiq Mobarak. “Effects of Emigration on Rural Labor Markets.” Working Paper, 2018. Angel, S (2011) Making room for a planet of cities. Policy Focus Report, Lincoln Institute of Land Policy, Alesina, Alberto, Sebastian Hohmann, Stelios Cambridge, MA. Michalopoulos, and Elias Papaioannou. “Intergen- erational Mobility in Africa.” Econometrica 89, Fno. 1 Arceo-Gomez, Eva, Rema Hanna, and Pauli- (2021): 1–35. https://doi.org/10.3982/ECTA17018. na Oliva. “Does the Effect of Pollution on Infant Mortality Differ Between Developing and Developed Alexander, Diane, and Hannes Schwandt. “The Countries? Evidence from Mexico City.” The Eco- Impact of Car Pollution on Infant and Child Health: nomic Journal 126, no. 591 (March 1, 2016): 257–80. Evidence from Emissions Cheating.” Working Paper. https://doi.org/10.1111/ecoj.12273. Federal Reserve Bank of Chicago, 2019. https://doi. org/10.21033/wp-2019-04. Ashraf, Nava, Edward Glaeser, Abraham Hol- land, and Bryce Millett Steinberg. “Water, Health Allen, Treb, and Costas Arkolakis. “Trade and the and Wealth.” Working Paper. National Bureau of Topography of the Spatial Economy.” The Quarterly Economic Research, September 2017. https://doi. Journal of Economics 129, no. 3 (August 1, 2014): org/10.3386/w23807. 1085–1140. https://doi.org/10.1093/qje/qju016.

Allen, Treb, Costas Arkolakis, and Xiangliang Li. “Optimal City Structure.” Working Paper, 2015.

EVIDENCE PAPER / CITIES 40 Ashraf, Nava, Edward L. Glaeser, and Giacomo Besley, Timothy, and Robin Burgess. “Land A. M. Ponzetto. “Infrastructure, Incentives, and Reform, Poverty Reduction, and Growth: Evidence Institutions.” American Economic Review 106, no. from India.” The Quarterly Journal of Economics 5 (May 2016): 77–82. https://doi.org/10.1257/aer. 115, no. 2 (May 1, 2000): 389–430. https://doi. p20161095. org/10.1162/003355300554809.

Autor, David H. “Work of the Past, Work of the Fu- Biagi, Bianca, and Claudio Detotto. “Crime as ture.” AEA Papers and Proceedings 109 (May 2019): Tourism Externality.” Regional Studies 48, no. 4 (April 1–32. https://doi.org/10.1257/pandp.20191110. 3, 2014): 693–709. https://doi.org/10.1080/0034340 4.2011.649005. Baruah, Neeraj G., Henderson, J. Vernon and Peng, Cong, “Colonial Legacies: Shaping African Bird, Julia, and Anthony J. Venables. “The Effi- Cities,” SERC Discussion Paper 226, 2017. ciency of Land-Use in a Developing City: Traditional vs Modern Tenure Systems in Kampala, Uganda.” Banerjee, Abhijit, Raghabendra Chattopadhyay, SSRN Scholarly Paper. Rochester, NY: Social Esther Duflo, Daniel Keniston, and Nina Singh. Science Research Network, March 1, 2019. http:// “Improving Police Performance in Rajasthan, India: papers.ssrn.com/abstract=3346354. Experimental Evidence on Incentives, Managerial Au- tonomy and Training.” American Economic Journal: Björkman, Martina, and Jakob Svensson. “Power Economic Policy, 13, no 1 (February 2021): 36-66. to the People: Evidence from a Randomised Field Ex- https://doi.org/10.1257/pol.20190664. periment on Community-Based Monitoring in Ugan- da.” The Quarterly Journal of Economics 124, no. Banerjee, Abhijit, Rema Hanna, and Sendhil 2 (May 1, 2009): 735–69. https://doi.org/10.1162/ Mullainathan. “Corruption.” In The Handbook of qjec.2009.124.2.735. Organizational Economics, Course Book. Prince- ton: Princeton University Press, 2013. https://doi. Blondel, Vincent D., Adeline Decuyper, and org/10.1515/9781400845354-029. Gautier Krings. “A Survey of Results on Mobile Phone Datasets Analysis.” EPJ Data Science 4, no. Banerjee, Abhijit V. “A Theory of Misgovern- 1 (December 2015): 1–55. https://doi.org/10.1140/ ance.” The Quarterly Journal of Economics 112, epjds/s13688-015-0046-0. no. 4 (November 1, 1997): 1289–1332. https://doi. org/10.1162/003355300555484. Brandily, P., and Ferdinand Gordian Rauch. “Roads and Urban Growth.” Working Paper, 2018. Barwick, Panle Jia, Dave Donaldson, Shanjun Li, and Yatang Lin. “The Welfare Effects of Passenger Brooks, Wyatt, and Kevin Donovan. “Eliminating Transportation Infrastructure: Evidence from China’s Uncertainty in Market Access: The Impact of New HSR Network.” Working Paper, 2019. Bridges in Rural Nicaragua.” Econometrica 88, no. 5 (2020): 1965–1997. https://doi.org/10.3982/ Baum-Snow, Nathaniel. “Did Highways Cause ECTA15828. Suburbanisation?” The Quarterly Journal of Econom- ics 122, no. 2 (May 1, 2007): 775–805. https://doi. Bryan, Gharad, Shyamal Chowdhury, and Ahmed org/10.1162/qjec.122.2.775. Mushfiq Mobarak. “Underinvestment in a Profit- able Technology: The Case of Seasonal Migration in Bangladesh.” Econometrica 82, no. 5 (2014): 1671–1748. https://doi.org/10.3982/ECTA10489.

EVIDENCE PAPER / CITIES 41 Bryan, Gharad, and Melanie Morten. “The Aggre- Chauvin, Juan Pablo, Edward Glaeser, Yueran gate Productivity Effects of Internal Migration: Evi- Ma, and Kristina Tobio. “What Is Different about dence from Indonesia.” Journal of Political Economy, Urbanisation in Rich and Poor Countries? Cities in 2019, 000–000. https://doi.org/10.1086/701810. Brazil, China, India and the United States.” Journal of Urban Economics, Urbanisation in Developing Buchmann, Nina, Erica M Field, Rachel Glen- Countries: Past and Present, 98 (March 1, 2017): nerster, and Reshmaan N Hussam. “Throwing 17–49. https://doi.org/10.1016/j.jue.2016.05.003. the Baby out with the Drinking Water: Unintended Consequences of Arsenic Mitigation Efforts in Bang- Chen, Martha and Beard, Victoria. “Including the ladesh.” Working Paper. National Bureau of Eco- Excluded: Supporting Informal Workers for More nomic Research, April 2019. https://doi.org/10.3386/ Equal and Productive Cities in the Global South” w25729. Working Paper. Washington DC. World Resources Institute. Calfee, John, and Clifford Winston. “The Value of Automobile Travel Time: Implications for Congestion Chetty, Raj, John N Friedman, Nathaniel Hen- Policy.” Journal of Public Economics 69, no. 1 (July dren, Maggie R Jones, and Sonya R Porter. “The 1, 1998): 83–102. https://doi.org/10.1016/S0047- Opportunity Atlas: Mapping the Childhood Roots 2727(97)00095-9. of Social Mobility.” Working Paper. National Bureau of Economic Research, October 2018. https://doi. Caliendo, Lorenzo, Maximiliano Dvorkin, and org/10.3386/w25147. Fernando Parro. “Trade and Labor Market Dynam- ics: General Equilibrium Analysis of the China Trade Chetty, Raj, and Nathaniel Hendren. “The Impacts Shock.” Econometrica 87, no. 3 (2019): 741–835. of Neighborhoods on Intergenerational Mobility I: https://doi.org/10.3982/ECTA13758. Childhood Exposure Effects.” The Quarterly Journal of Economics 133, no. 3 (August 1, 2018): 1107–62. Cesur, Resul, Erdal Tekin, and Aydogan Ulker. https://doi.org/10.1093/qje/qjy007. “Air Pollution and Infant Mortality: Evidence from the Expansion of Natural Gas Infrastructure.” The Eco- Chetty, Raj, Nathaniel Hendren, and Lawrence F. nomic Journal 127, no. 600 (March 1, 2017): 330–62. Katz. “The Effects of Exposure to Better Neighbor- https://doi.org/10.1111/ecoj.12285. hoods on Children: New Evidence from the Moving to Opportunity Experiment.” American Economic ———. “Can Natural Gas Save Lives? Evidence Review 106, no. 4 (April 2016): 855–902. https://doi. from the Deployment of a Fuel Delivery System in a org/10.1257/aer.20150572. Developing Country.” Journal of Health Economics 59 (May 2018): 91–108. https://doi.org/10.1016/j. Ciccone, Antonio, and Robert E. Hall. “Produc- jhealeco.2018.03.001. tivity and the Density of Economic Activity.” The American Economic Review 86, no. 1 (1996): 54–70. Chalfin, Aaron. “Economic Costs of Crime.” In The Encyclopedia of Crime & Punishment, 1–12. Coase, R. H. “The Problem of Social Cost.” The American Cancer Society, 2015. https://doi. Journal of Law & Economics 3 (1960): 1–44. org/10.1002/9781118519639.wbecpx193. Collier, P., Glaeser, E., Venables, A., Haas, A., Wani, S. “Designed to Succeed: Building an Author- ising Environment for Fast-Growing Cities.” Interna- tional Growth Centre (forthcoming).

EVIDENCE PAPER / CITIES 42 Combes, Pierre-Philippe, Sylvie Démurger, Shi Davis, Lucas W. “The Effect of Driving Restrictions Li, and Jianguo Wang. “Unequal Migration and on Air Quality in Mexico City.” Journal of Political Urbanisation Gains in China.” Journal of Devel- Economy 116, no. 1 (February 1, 2008): 38–81. opment Economics, January 29, 2019. https://doi. https://doi.org/10.1086/529398. org/10.1016/j.jdeveco.2019.01.009. De La Roca, Jorge, and Diego Puga. “Learning Combes, Pierre-Philippe, Gilles Duranton, Lau- by Working in Big Cities.” The Review of Economic rent Gobillon, and Sébastien Roux. “Estimating Studies 84, no. 1 (January 1, 2017): 106–42. https:// Agglomeration Economies with History, Geology, doi.org/10.1093/restud/rdw031. and Worker Effects.” In Agglomeration Economics, edited by Edward L. Glaeser, 15–66. The University Desmet, Klaus, and Esteban Rossi-Hansberg. of Chicago Press, 2010. http://www.nber.org/chap- “On the Spatial Economic Impact of Global Warm- ters/c7978. ing.” Journal of Urban Economics 88 (July 2015): 16–37. https://doi.org/10.1016/j.jue.2015.04.004. Couture, Victor, Cecile Gaubert, Jessie Hand- bury, and Erik Hurst. “Income Growth and the Devoto, Florencia, Esther Duflo, Pascaline Dupas, Distributional Effects of Urban Spatial Sorting,” William Parienté, and Vincent Pons. “Happiness 2019, 87. on Tap: Piped Water Adoption in Urban Morocco.” American Economic Journal: Economic Policy 4, Cullen, Julie Berry, and Steven D. Levitt. “Crime, no. 4 (May 2012): 68–99. https://doi.org/10.1257/ Urban Flight, And The Consequences For Cities.” pol.4.4.68. The Review of Economics and Statistics 81, no. 2 (1999): 159–69. Djankov, Simeon, Rafael La Porta, Floren- cio Lopez-de-Silanes, and Andrei Shleifer. Currie, Janet, Matthew Neidell, and Johannes F. “Courts.” The Quarterly Journal of Economics Schmieder. “Air Pollution and Infant Health: Les- 118, no. 2 (May 1, 2003): 453–517. https://doi. sons from New Jersey.” Journal of Health Econom- org/10.1162/003355303321675437. ics 28, no. 3 (May 1, 2009): 688–703. https://doi. org/10.1016/j.jhealeco.2009.02.001. Donaldson, Dave. “Railroads of the Raj: Estimat- ing the Impact of Transportation Infrastructure.” Currie, Janet, and Tom Vogl. “Early-Life Health American Economic Review 108, no. 4–5 (April 2018): and Adult Circumstance in Developing Coun- 899–934. https://doi.org/10.1257/aer.20101199. tries.” Annual Review of Economics 5, no. 1 (2013): 1–36. https://doi.org/10.1146/annurev-econom- Donaldson, Dave, and Richard Hornbeck. “Rail- ics-081412-103704. roads and American Economic Growth: A ‘Market Access’ Approach.” The Quarterly Journal of Eco- Cutler, David, and Grant Miller. “The Role of nomics 131, no. 2 (May 1, 2016): 799–858. https:// Public Health Improvements in Health Advances: doi.org/10.1093/qje/qjw002. The Twentieth-Century United States.” Demogra- phy 42, no. 1 (February 1, 2005): 1–22. https://doi. Duranton, Gilles, and Matthew A. Turner. “The org/10.1353/dem.2005.0002. Fundamental Law of Road Congestion: Evidence from US Cities.” American Economic Review 101, no. 6 (October 2011): 2616–52. https://doi. org/10.1257/aer.101.6.2616.

EVIDENCE PAPER / CITIES 43 ———. “Urban Growth and Transportation.” The Re- Ferraz, Claudio, and Frederico Finan. “Electoral view of Economic Studies 79, no. 4 (October 2012): Accountability and Corruption: Evidence from the 1407–40. https://doi.org/10.1093/restud/rds010. Audits of Local Governments.” American Economic Review 101, no. 4 (June 2011): 1274–1311. https:// Edin, Per-Anders, Peter Fredriksson, and Olof doi.org/10.1257/aer.101.4.1274. Åslund. “Settlement Policies and the Economic Suc- cess of Immigrants.” Journal of Population Econom- Field, Erica. “Entitled to Work: Urban Property ics 17, no. 1 (February 1, 2004): 133–55. https://doi. Rights and Labor Supply in Peru.” The Quarter- org/10.1007/s00148-003-0143-4. ly Journal of Economics 122, no. 4 (November 1, 2007): 1561–1602. https://doi.org/10.1162/ Ehrlich, Isaac. “The Deterrent Effect of Capital Pun- qjec.2007.122.4.1561. ishment: A Question of Life and Death.” The Ameri- can Economic Review 65, no. 3 (1975): 397–417. Fogli, Alessandra, and Veronica Guerrieri. “The End of the American Dream? Inequality and Segre- Ellison, Glenn, Edward L. Glaeser, and William R. gation in US Cities.” Working Paper. National Bureau Kerr. “What Causes Industry Agglomeration? Evi- of Economic Research, August 2019. https://doi. dence from Coagglomeration Patterns.” American org/10.3386/w26143. Economic Review 100, no. 3 (June 2010): 1195–1213. https://doi.org/10.1257/aer.100.3.1195. Franklin, Simon. “The Demand for Government Housing: Evidence from Lotteries for 200,000 Homes Engel, Eduardo, Ronald D. Fischer, and Alexan- in Ethiopia.” Working Paper, 2019. der Galetovic P. The Economics of Public-Private Partnerships: A Basic Guide. New York, NY: Cam- Gamper-Rabindran, Shanti, Shakeeb Khan, and bridge University Press, 2014. Christopher Timmins. “The Impact of Piped Water Provision on Infant Mortality in Brazil: A Quantile Faber, Benjamin, and Cecile Gaubert. “Tourism Panel Data Approach.” Journal of Development and Economic Development: Evidence from Mex- Economics 92, no. 2 (July 1, 2010): 188–200. https:// ico’s Coastline.” American Economic Review 109, doi.org/10.1016/j.jdeveco.2009.02.006. no. 6 (June 2019): 2245–93. https://doi.org/10.1257/ aer.20161434. Gechter, Michael, and Nick Tsivanidis. “The Welfare Consequences of Formalizing Developing Fajgelbaum, Pablo D., Eduardo Morales, Juan Country Cities: Evidence from the Mumbai Mills Carlos Suárez Serrato, and Owen Zidar. “State Redevelopment.” Working Paper, 2018. Taxes and Spatial Misallocation.” The Review of Economic Studies 86, no. 1 (January 1, 2019): Giusti, L. “A Review of Waste Management Prac- 333–76. https://doi.org/10.1093/restud/rdy050. tices and Their Impact on Human Health.” Waste Management (New York, N.Y.) 29, no. 8 (August Fajgelbaum, Pablo D, and Edouard Schaal. 2009): 2227–39. https://doi.org/10.1016/j.was- “Optimal Transport Networks in Spatial Equilibri- man.2009.03.028. um.” Working Paper. National Bureau of Economic Research, 2019. https://doi.org/10.3386/w23200. Glaeser, Edward L., and Joshua Gottlieb. “The Economics of Place-Making Policies.” Brookings Pa- Fan, Mingxuan, and Corbett Grainger. “The pers on Economic Activity 39, no. 1 (Spring) (2008): Impact of Air Pollution on Labor Supply in China.” 155–253. Working Paper, 2019.

EVIDENCE PAPER / CITIES 44 Glaeser, Edward L., and Joshua D. Gottlieb. “The Gollin, Douglas, Martina Kirchberger, and David Wealth of Cities: Agglomeration Economies and Lagakos. “In Search of a Spatial Equilibrium in the Spatial Equilibrium in the United States.” Journal Developing World.” Working Paper. National Bureau of Economic Literature 47, no. 4 (December 2009): of Economic Research, October 2017. https://doi. 983–1028. https://doi.org/10.1257/jel.47.4.983. org/10.3386/w23916.

Glaeser, Edward L., and Matthew E. Kahn. “The Gollin, Douglas, David Lagakos, and Michael E. Greenness of Cities: Carbon Dioxide Emissions Waugh. “The Agricultural Productivity Gap.” The and Urban Development.” Journal of Urban Eco- Quarterly Journal of Economics 129, no. 2 (May 1, nomics 67, no. 3 (May 1, 2010): 404–18. https://doi. 2014): 939–93. https://doi.org/10.1093/qje/qjt056. org/10.1016/j.jue.2009.11.006. Gonzalez-Navarro, Marco, and Clement Quin- Glaeser, Edward L., Sari Pekkala Kerr, and Wil- tana-Domeque. “Paving Streets for the Poor: liam R. Kerr. “Entrepreneurship and Urban Growth: Experimental Analysis of Infrastructure Effects.” The An Empirical Assessment with Historical Mines.” Review of Economics and Statistics 98, no. 2 (No- Review of Economics and Statistics, April 23, 2015. vember 16, 2015): 254–67. https://doi.org/10.1162/ https://doi.org/10.1162/REST_a_00456. REST_a_00553.

Glaeser, Edward L., and Matthew G. Resseger. Gramlich, Edward M. “Infrastructure Investment: “The Complementarity Between Cities and Skills*.” A Review Essay.” Journal of Economic Literature 32, Journal of Regional Science 50, no. 1 (2010): 221–44. no. 3 (1994): 1176–96. https://doi.org/10.1111/j.1467-9787.2009.00635.x. Greenstone, Michael, Richard Hornbeck, and Glaeser, Edward L., Joseph A Scheinkman, and Enrico Moretti. “Identifying Agglomeration Spillo- Andrei Shleifer. “Economic Growth in a Cross-Sec- vers: Evidence from Winners and Losers of Large tion of Cities.” Journal of Monetary Economics 36, Plant Openings.” Journal of Political Economy no. 1 (1995): 117–43. 118, no. 3 (June 1, 2010): 536–98. https://doi. org/10.1086/653714. Glaeser, Edward L., and David C. Maré. “Cities and Skills.” Journal of Labor Econom- Guimarães, Paulo, Octávio Figueiredo, and ics 19, no. 2 (April 1, 2001): 316–42. https://doi. Douglas Woodward. “Agglomeration and the Loca- org/10.1086/319563. tion of Foreign Direct Investment in Portugal.” Jour- nal of Urban Economics 47, no. 1 (January 1, 2000): Goldstein, Markus, and Christopher Udry. 115–35. https://doi.org/10.1006/juec.1999.2138. “The Profits of Power: Land Rights and Agricul- tural Investment in Ghana.” Journal of Political Hanna, Rema, Gabriel Kreindler, and Benjamin Economy 116, no. 6 (2008): 981–1022. https://doi. A. Olken. “Citywide Effects of High-Occupancy org/10.1086/595561. Vehicle Restrictions: Evidence from ‘Three-in-One’ in Jakarta.” Science 357, no. 6346 (July 7, 2017): Gollin, Douglas, Remi Jedwab, and Dietrich Vol- 89–93. https://doi.org/10.1126/science.aan2747. lrath. “Urbanisation with and without Industrialisa- tion.” Journal of Economic Growth 21, no. 1 (2016): 35–70.

EVIDENCE PAPER / CITIES 45 Hanna, Rema, and Paulina Oliva. “The Effect of Hicks, Joan Hamory, Marieke Kleemans, Nicholas Pollution on Labor Supply: Evidence from a Natural Y Li, and Edward Miguel. “Reevaluating Agricultur- Experiment in Mexico City.” Journal of Public Eco- al Productivity Gaps with Longitudinal Microdata.” nomics 122 (February 1, 2015): 68–79. https://doi. Journal of the European Economic Association, org/10.1016/j.jpubeco.2014.10.004. (November 2020). https://doi.org/10.1093/jeea/ jvaa043. Harari, Mariaflavia. “Cities in Bad Shape: Urban Geometry in India.” American Economic Review Hurwicz, Leonid. “But Who Will Guard the Guard- 110, no 8 (August 2020): 2377-2421. https://doi. ians?” American Economic Review 98, no. 3 (June org/10.1257/aer.20171673. 2008): 577–85. https://doi.org/10.1257/aer.98.3.577.

Harari, Mariaflavia, and Maisy Wong. “Slum Imbert, Clement, Marlon Seror, Yifan Zhang, and Upgrading and Long-Run Urban Development: Yanos Zylberberg. “Migrants and Firms: Evidence Evidence from Indonesia.” 2018 Meeting Papers. from China.” Working Paper. CESifo Working Paper Society for Economic Dynamics, 2018. https://ideas. Series. CESifo Group Munich, 2018. https://ideas. repec.org/p/red/sed018/367.html. repec.org/p/ces/ceswps/_7440.html.

Heath, Rachel, Ghazala Mansuri, and Bob Rijk- Jaffe, Adam B., Manuel Trajtenberg, and Re- ers. “Labor Supply Responses to Illness Shocks: becca Henderson. “Geographic Localization Evidence from High-Frequency Labor Market Data of Knowledge Spillovers as Evidenced by Patent from Urban Ghana.” Journal of Human Resources, Citations.” The Quarterly Journal of Economics 0618-9584R2. 108, no. 3 (August 1, 1993): 577–98. https://doi. org/10.2307/2118401. Heblich, Stephan, Stephen J Redding, and Daniel M Sturm. “The Making of the Modern Metropolis: Kahn, Matthew E. Climatopolis: How Our Cities Evidence from London.” The Quarterly Journal of Will Thrive in the Hotter Future. 1 edition. New York: Economics 135, no. 4 (November 2020): 2059–2133. Basic Books, 2010. https://doi.org/10.1093/qje/qjaa014. ———. “The Silver Lining of Rust Belt Manufacturing Henderson, J. Vernon. “Marshall’s Economies.” Decline.” Journal of Urban Economics 46, no. 3 (No- Journal of Urban Economics 53, no. 1 (January, vember 1, 1999): 360–76. https://doi.org/10.1006/ 2003): 1–28. https://doi.org/10.1016/S0094- juec.1998.2127. 1190(02)00505-3. Kain, John F. “The Use of Straw Men in the Eco- ———. “The Effects of Urban Concentration on nomic Evaluation of Rail Transport Projects.” The Economic Growth.” Working Paper. National Bureau American Economic Review 82, no. 2 (1992): 487–93. of Economic Research, January 2000. https://doi. org/10.3386/w7503. Katz, Lawrence F., Jeffrey R. Kling, and Jef- frey B. Liebman. “Moving to Opportunity in Henderson, J. Vernon, Tanner Regan, and Antho- Boston: Early Results of a Randomised Mobility ny J. Venables. “Building the City: Urban Transition Experiment.” The Quarterly Journal of Econom- and Institutional Frictions.” SERC Discussion Pa- ics 116, no. 2 (May 1, 2001): 607–54. https://doi. pers. Spatial Economics Research Centre, LSE, April org/10.1162/00335530151144113. 2016. https://ideas.repec.org/p/cep/sercdp/0196. html.

EVIDENCE PAPER / CITIES 46 Kessler, Daniel, and Steven D. Levitt. “Using ———. “Using Electoral Cycles in Police Hiring to Sentence Enhancements to Distinguish between Estimate the Effect of Police on Crime.” The Ameri- Deterrence and Incapacitation.” The Journal of Law can Economic Review 87, no. 3 (1997): 270–90. and Economics 42, no. S1 (April 1, 1999): 343–64. https://doi.org/10.1086/467428. Lucas, Robert E. “On the Mechanics of Eco- nomic Development.” Journal of Monetary Eco- Khan, Adnan Q., Asim I. Khwaja, and Benjamin A. nomics 22, no. 1 (July 1, 1988): 3–42. https://doi. Olken. “Tax Farming Redux: Experimental Evidence org/10.1016/0304-3932(88)90168-7. on Performance Pay for Tax Collectors.” The Quar- terly Journal of Economics 131, no. 1 (February 1, Ludwig, Jens, and Philip J. Cook. “The Benefits 2016): 219–71. https://doi.org/10.1093/qje/qjv042. of Reducing Gun Violence: Evidence from Contin- gent-Valuation Survey Data.” Journal of Risk and Kreindler, Gabriel. “Driving Delhi? Behavioural Uncertainty 22, no. 3 (2001): 207–26. Responses to Driving Restrictions.” Working Paper, 2016. https://www.ssrn.com/abstract=2966797. Marx, Benjamin, Thomas Stoker, and Tavneet Suri. “The Economics of Slums in the Developing Kreindler, Gabriel E. “The Welfare Effect of Road World.” Journal of Economic Perspectives 27, no. 4 Congestion Pricing: Experimental Evidence and Equi- (November 2013): 187–210. https://doi.org/10.1257/ librium Implications.” Working Paper, 2018. jep.27.4.187.

Kreindler, Gabriel E, and Yuhei Miyauchi. “Meas- Meyer, J. R., J. F. Kain, and M. Wohl. The Urban uring Commuting and Economic Activity inside Transportation Problem. Originally published 1965; Cities with Cell Phone Records,” 2019, 47. Im Original erschienen 1965. Cambridge, MA and London, England: Harvard University Press, 1965. Krugman, Paul. “Increasing Returns and Eco- nomic Geography.” Journal of Political Econo- Michaels, Guy, Dzhamilya Nigmatulina, Ferdi- my 99, no. 3 (June 1, 1991): 483–99. https://doi. nand Rauch, Tanner Regan, Neeraj Baruah, and org/10.1086/261763. Amanda Dahlstrand-Rudin. “Planning Ahead for Better Neighborhoods: Long Run Evidence from Lagakos, David, and Michael E. Waugh. “Selec- Tanzania.” Journal of Political Economy. https://doi. tion, Agriculture, and Cross-Country Productivity org/10.1086/714119. Differences.” American Economic Review 103, no. 2 (April 2013): 948–80. https://doi.org/10.1257/ Mills, Edwin S. “An Aggregative Model of Resource aer.103.2.948. Allocation in a Metropolitan Area.” The American Economic Review 57, no. 2 (1967): 197–210. Leape, Jonathan. “The London Congestion Charge.” Journal of Economic Perspectives 20, no. 4 Misse, Michel, and Joana Vargas. “O Fluxo Do (December 2006): 157–76. https://doi.org/10.1257/ Processo de Incriminação No Rio de Janeiro Na jep.20.4.157. Década de 50 e No Período de 1997-2001: Com- paração e Análise.” In 13o Congresso Brasileiro de Levitt, Steven D. “Juvenile Crime and Punishment.” Sociologia, Recife, PE, 2007. Journal of Political Economy 106, no. 6 (1998): 1156–85. https://doi.org/10.1086/250043.

EVIDENCE PAPER / CITIES 47 Monte, Ferdinando, Stephen J. Redding, and Nakamura, Emi, Jósef Sigurdsson, and Jón Esteban Rossi-Hansberg. “Commuting, Migra- Steinsson. “The Gift of Moving: Intergenerational tion, and Local Employment Elasticities.” American Consequences of a Mobility Shock.” Working Paper. Economic Review 108, no. 12 (December 2018): National Bureau of Economic Research, July 2016. 3855–90. https://doi.org/10.1257/aer.20151507. https://doi.org/10.3386/w22392.

Moretti, Enrico. “Estimating the Social Return to Owens III, Raymond, Esteban Rossi-Hansberg, Higher Education: Evidence from Longitudinal and and Pierre-Daniel Sarte. “Rethinking Detroit.” Repeated Cross-Sectional Data.” Journal of Econo- American Economic Journal: Economic Policy, 2019. metrics, Higher education (Annals issue), 121, no. https://doi.org/10.1257/pol.20180651. 1 (July 1, 2004): 175–212. https://doi.org/10.1016/j. jeconom.2003.10.015. OECD (2015), The Metropolitan Century: Understanding Urbanisation and its Conse- Muralidharan, Karthik, and Venkatesh Sundarar- quences, OECD Publishing, Paris, https://doi. aman. “Teacher Performance Pay: Experimental Evi- org/10.1787/9789264228733-en. dence from India.” Journal of Political Economy 119, no. 1 (2011): 39–77. https://doi.org/10.1086/659655. Perlman, Janice. Favela: Four Decades of Living on the Edge in Rio de Janeiro. New York, N.Y.: Oxford ———. “The Aggregate Effect of School Choice: Ev- University Press, 2010. idence from a Two-Stage Experiment in India.” The Quarterly Journal of Economics 130, no. 3 (August 1, Plato. The Republic. Translated by Alexander Kerr. 2015): 1011–66. https://doi.org/10.1093/qje/qjv013. Stilwell: Neeland Media LLC, 2016.

Muth, R. F. Cities and Housing the Spatial Pattern Psacharopoulos, George, and Harry Anthony of Urban Residential Land Use. The University of Patrinos. “Returns to Investment in Education : A Chicago Press, 1969. Decennial Review of the Global Literature.” Educa- tion Economics 26, no 5, (June 2018): 445-458. DOI: Nagin, Daniel S. “Deterrence: A Review of the https://doi.org/10.1080/09645292.2018.1484426. Evidence by a Criminologist for Economists.” Annual Review of Economics 5, no. 1 (2013): Quah, Euston, and Tay Liam Boon. “The Eco- 83–105. https://doi.org/10.1146/annurev-econom- nomic Cost of Particulate Air Pollution on Health in ics-072412-131310. Singapore.” Journal of Asian Economics 14, no. 1 (February 1, 2003): 73–90. https://doi.org/10.1016/ Naik, Nikhil, Scott Duke Kominers, Ramesh Ras- S1049-0078(02)00240-3. kar, Edward L Glaeser, and César A Hidalgo. “Do People Shape Cities, or Do Cities Shape People? Rauch, James E. “Productivity Gains from Geo- The Co-Evolution of Physical, Social, and Economic graphic Concentration of Human Capital: Evidence Change in Five Major U.S. Cities.” Working Paper. from the Cities.” Journal of Urban Economics 34, National Bureau of Economic Research, October no. 3 (November 1, 1993): 380–400. https://doi. 2015. https://doi.org/10.3386/w21620. org/10.1006/juec.1993.1042.

Redding, Stephen J. “Goods Trade, Factor Mobility and Welfare.” Journal of International Economics 101 (July 1, 2016): 148–67. https://doi.org/10.1016/j. jinteco.2016.04.003.

EVIDENCE PAPER / CITIES 48 Redding, Stephen J., and Esteban Rossi-Hans- Troesken, Werner. “Lead Water Pipes and Infant berg. “Quantitative Spatial Economics.” Annual Mortality at the Turn of the Twentieth Century.” Review of Economics 9, no. 1 (2017): 21–58. https:// Journal of Human Resources 43, no. 3 (July 1, 2008): doi.org/10.1146/annurev-economics-063016-103713. 553–75. https://doi.org/10.3368/jhr.43.3.553.

Rosenstein-Rodan, P. N. “Problems of Industrial- Tsivanidis, Nick. “The Aggregate and Distributional isation of Eastern and South-Eastern Europe.” The Effects of Urban Transit Infrastructure: Evidence from Economic Journal 53, no. 210/211 (1943): 202–11. Bogotá’s TransMilenio,” 2019, 62. https://doi.org/10.2307/2226317. United Nations. “2018 Revision of World Urbani- Ellis, P. and Roberts, M. “Leveraging urbanisation sation Prospects.” Press Release. United Nations in South Asia: Managing spatial transformation for Department of Economic and Social Affairs, May 16, prosperity and livability.” The World Bank (2015). 2018. https://www.un.org/development/desa/publications/2018-revision-of-world-urbanisation-pros- Sarvimäki, Matti, Roope Uusitalo, and Markus pects.html. Jäntti. “Habit Formation and the Misallocation of La- bor: Evidence from Forced Migrations.” Working Pa- Vogel, Kathryn Baragwanath, Ran Goldblatt, per, 2019. https://www.ssrn.com/abstract=3361356. Gordon H Hanson, and Amit K Khandelwal. “Detecting Urban Markets with Satellite Imagery: Singh, Ram. “Public–Private Partnerships vs. Tra- An Application to India.” Journal of Urban Eco- ditional Contracts for Highways.” Indian Economic nomics, (25 June 2019). https://doi.org/10.1016/j. Review 53, no. 1 (December 1, 2018): 29–63. https:// jue.2019.05.004. doi.org/10.1007/s41775-018-0032-0. Vollrath, Dietrich. “Agriculture and Growth Soto, Hernando de. The Mystery of Capital: Why Reading List.” Growth Economics (blog), May Capitalism Triumphs in the West and Fails Every- 18, 2014. https://growthecon.com/blog/agricul- where Else. 1st pbk. ed. New York: Basic Books, ture-and-growth-reading-list/. 2000. World Justice Project. The World Justice Project: Suzuki, Hiroaki, Jin Murakami, Yu-Hung Hong, Rule of Law Index, 2019, 2019. and Beth Tamayose. Financing Transit-Oriented Development with Land Values: Adapting Land Value Young, Alwyn. “Inequality, the Urban-Rural Gap, Capture in Developing Countries. Washington, DC: and Migration*.” The Quarterly Journal of Econom- World Bank Publications, 2015. ics 128, no. 4 (November 1, 2013): 1727–85. https:// doi.org/10.1093/qje/qjt025. Thaler, Richard, and Sherwin Rosen. “The Value of Saving a Life: Evidence from the Labor Market.” In Household Production and Consumption, edited by Nestor E. Terleckyj, 265–302, 1976. https://www. nber.org/chapters/c3964.

EVIDENCE PAPER / CITIES 49 theigc.org DIRECTED BY FUNDED BY